KEY FORMULAS
Prem S. Mann • Introductory Statistics, Seventh Edition
Chapter 2 • Organizing and Graphing Data
• Relative frequency of a class = //X/
• Percentage of a class = (Relative frequency) X 100
• Class midpoint or mark = (Upper limit + Lower limit)/2
• Class width = Upper boundary — Lower boundary
• Cumulative relative frequency
Cumulative frequency
Total observations in the data set
• Cumulative percentage
= (Cumulative relative frequency) X 100
Chapter 3 • Numerical Descriptive Measures
• Mean for ungrouped data: jj, = %x/N and x = %x/n
• Mean for grouped data: fi = Xmf/N and x = Xmf/n
where m is the midpoint and /is the frequency of a class
• Median for ungrouped data
= Value of the middle term in a ranked data set
Range = Largest value — Smallest value
Variance for ungrouped data:
Ix 2 -
(X*) 2 ^ 2 (X*) 2
N
X* 2 -
and s
N n - 1
where a 2 is the population variance and s 2 is the sample
variance
Standard deviation for ungrouped data:
Xx 2
(X*) 2
N
X* 2 -
(X*) 2
and s -
N V n - 1
where a and s are the population and sample standard de-
viations, respectively
Variance for grouped data:
Xm 2 /-
(Xm/) 2
N
Xm 2 f-
(Xm/) 2
a~
N
and s
n - 1
Standard deviation for grouped data:
(Xmff
Xm 2 f-
(2mff
N
Zm 2 f-
N
and s
n - 1
Chebyshev's theorem:
For any number k greater than 1, at least (1 — 1/A: 2 ) of the
values for any distribution lie within k standard deviations
of the mean.
Empirical rule:
For a specific bell-shaped distribution, about 68% of the ob-
servations fall in the interval (/i — cr) to (fi + cr), about
95% fall in the interval {pi — 2a) to (/j, + 2a), and about
99.7% fall in the interval (ji - 3a) to (fi + 3a).
Q x = First quartile given by the value of the middle term
among the (ranked) observations that are less than the
median
Q 2 = Second quartile given by the value of the middle term
in a ranked data set
Q 3 = Third quartile given by the value of the middle term
among the (ranked) observations that are greater than
the median
Interquartile range:
The &th percentile:
P t = Value of the
IQR = Qs-Qi
kn \
100
jth term in a ranked data set
Percentile rank of x,
Number of values less than x,
Total number of values in the data set
X 100
Chapter 4 • Probability
• Classical probability rule for a simple event:
P(Ej) =
Total number of outcomes
• Classical probability rule for a compound event:
. , Number of outcomes in A
P(A) =
Total number of outcomes
• Relative frequency as an approximation of probability:
P(A) = f -
n
• Conditional probability of an event:
P(A and B) P(A and B)
• Condition for independence of events:
P(A) = P(A\B) and/or P(B) = P(B\A)
• For complementary events: P(A) + P(A) = 1
• Multiplication rule for dependent events:
P(A andfi) = P(A) P(B\A)
• Multiplication rule for independent events:
P(A and B) = P(A) P(B)
• Joint probability of two mutually exclusive events:
P(A and B) =
• Addition rule for mutually nonexclusive events:
P(A or B) = P(A) + P(B) - P(A and B)
• Addition rule for mutually exclusive events:
P(A or B) = P(A) + P(B)
Population proportion: p = X/N
Sample proportion: p = x/n
Mean of p: pp = p
Standard deviation of p when n/N ^ .05: cr~ = \fpqjn
p — p
z value for p: z =
Chapter 5 * Discrete Random Variables and Their
Probability Distributions
• Mean of a discrete random variable x: p. = ~XxP(x)
• Standard deviation of a discrete random variable x:
cr = VSx 2 P(x) - p 2
• n factorial: n\ = n(n — l)(n — 2) ... 3 • 2 • 1
• Number of combinations of n items selected x at a time:
r
x\{n — x)\
Number of permutations of n items selected x at a time:
n\
" x (« - x)\
• Binomial probability formula: P(x) = n C x p x q"~ x
• Mean and standard deviation of the binomial distribution:
p = np and cr = \fnpq
• Hypergeometric probability formula:
fix N-fin-x
P( X )
nC„
Poisson probability formula: P(x)
K x e
Mean, variance, and standard deviation of the Poisson prob-
ability distribution:
p = A, cr 2 = A, and cr = VX
Chapter 6 • Continuous Random Variables
and the Normal Distribution
z value for an x value: z
x — p
• Value of x when p, cr, and z are known: x = p + zcr
Chapter 7 • Sampling Distributions
• Mean of x : p- x = p
• Standard deviation of x when n/N ^ .05: <x T = cr/wn
_ x — p
• z value for x: z =
(Tt
Chapter 8 • Estimation of the Mean and Proportion
• Point estimate of p = x
• Confidence interval for p using the normal distribution
when cr is known:
x ± zcr- where cr s = a/vn
• Confidence interval for p using the t distribution when cr is
not known:
x ± ts x where s x = s/vn
• Margin of error of the estimate for p:
E = zcr- x or ts- x
• Determining sample size for estimating p:
n = z 2 cr 2 /E 2
• Confidence interval for p for a large sample:
p ± zsp where Sp = \/pq/ n
Margin of error of the estimate for p:
E = zsp, where Sp = 'Vpq/n
Determining sample size for estimating p:
n = z 2 pq/E 2
Chapter 9 • Hypothesis Tests about the Mean
and Proportion
• Test statistic z for a test of hypothesis about p using the
normal distribution when cr is known:
x - p cr
z = where ctt: =
Test statistic for a test of hypothesis about p using the t dis-
tribution when cr is not known:
x - p s
t = where = — 1=
s x Vn
Test statistic for a test of hypothesis about p for a large
sample:
P ~ P
z = where cr fl =
Chapter 10 * Estimation and Hypothesis Testing:
Two Populations
• Mean of the sampling distribution of x x — x 2 .
= Ml _ M*2
• Confidence interval for p, x — /x 2 for two independent
samples using the normal distribution when cr, and cr 2 are
known:
la] a\
\*\ ~ x 2> ± Z°Vi 2 where ^-i, = \\— + —
Test statistic for a test of hypothesis about /a, — /a 2 for two
independent samples using the normal distribution when cr,
and £r 2 are known:
(ii - x 2 ) - (mi - M2)
z =
For two independent samples taken from two populations
with equal but unknown standard deviations:
Pooled standard deviation:
(n x - l)sf + (n 2 - 1)4
V »i + n 2 — 2
Estimate of the standard deviation of x x — x 2 :
1 1
— + —
n x n 2
Confidence interval for j± x — /jl 2 using the t distribution:
(x x - x 2 ) ± ts 7xi ^
Test statistic using the t distribution:
t = {x\ ~ xi) ~ (Mi ~ M2)
For two independent samples selected from two populations
with unequal and unknown standard deviations:
'2 2\2
s i + fi)
Degrees of freedom: df = / 2 . 2 , 2 . 2
■12
Estimate of the standard deviation of x x — x 2 .
~2 2
n, n 2
Confidence interval for fi l — /jl 2 using the t distribution:
(jcj - x 2 ) ± ts- x ^- Xi
Test statistic using the t distribution:
(jc, - x 2 ) ~ (Mi - M2)
For two paired or matched samples:
Sample mean for paired differences: d = Xd/n
Sample standard deviation for paired differences:
Xd
V n — 1
Mean and standard deviation of the sampling distribution
of d:
MS = Mrf and s~ d = s d /Vn
Confidence interval for /j, d using the t distribution:
d ± ts d where s~, = s d /vn
Test statistic for a test of hypothesis about jj, d using the t
distribution:
For two large and independent samples, confidence interval
for pi - p 2 :
PiQi P2Q2
(Pi ~ P2) ± zs Pl -p 2 where s A _ A = yj — + —
For two large and independent samples, for a test of
hypothesis about p x — p 2 with H : p x — p 2 = 0:
Pooled sample proportion:
_ x x + x 2 n x p x + n 2 p 2
p = — — — or
«i + n 2 n x + n 2
Estimate of the standard deviation of p x — p 2 .
\ — + —
1 n x n 2
Test statistic: z
{P\ ~ P2) ~ (Pi ~ P2)
Chapter 1 1 • Chi-Square Tests
• Expected frequency for a category for a goodness-of-fit
test:
E = np
• Degrees of freedom for a goodness-of-fit test:
df = k — 1 where k is the number of categories
• Expected frequency for a cell for an independence or ho-
mogeneity test:
(Row total)(Column total)
E =
Sample size
• Degrees of freedom for a test of independence or
homogeneity:
df=(R- 1)(C- 1)
where R and C are the total number of rows and columns,
respectively, in the contingency table
Test statistic for a goodness-of-fit test and a test of inde-
pendence or homogeneity:
x 2 = V-
Confidence interval for the population variance a 2
(re - \)s 2 (re - l)s 2
2 to 2
Xa/2 XT -a/2
Test statistic for a test of hypothesis about a 2 :
2 in - iy
X' = ~2
Chapter 12 • Analysis of Variance
Let:
k = the number of different samples
(or treatments)
the size of sample i
the sum of the values in sample ;
the number of values in all samples
n \ + n 2 + n 3 + ' ' '
the sum of the values in all samples
r, + t 2 + t 3 + ■ ■ ■
the sum of the squares of values in all samples
For the F distribution:
Degrees of freedom for the numerator = k — 1
Degrees of freedom for the denominator = n — k
Between-samples sum of squares:
T,
n
Zx
2* 2
SSB
+
+
3 | ) (M
Within-samples sum of squares:
SSW = Xx 2
+
+ — + •■
"3
Sum of squares of xy, xx, and yy:
(Xx)(Sy)
SS„ = Xxy
ss,.
(Xx)
and SS,
Xy :
(Xy) 2
Least squares estimates of A and B:
b = SS„,/SS xt and a = y — bx
Standard deviation of the sample errors:
S S yy b S S ™
re - 2
• Error sum of squares: SSE = Se 2 = X(y — y) 2
(Sy) 2
• Total sum of squares: SST = 2y 2
n
• Regression sum of squares: SSR = SST — SSE
• Coefficient of determination: r 2 = b SS rv /SS vv
• Confidence interval for B:
b ± ts b where s h = s f /VSS„
• Test statistic for a test of hypothesis about B: t =
S S A - V
• Linear correlation coefficient: r = , =
SS VV
• Test statistic for a test of hypothesis about p:
B
1
Confidence interval for /jb y \ x :
y ± tss where sa
Prediction interval for y p :
y ± ts<s where s$ =
1 (x -x) 2
1 + I + ^ ~ X Y
n SS„
(Xx)
• Total sum of squares:
SST = SSB + SSW = %:
n
• Variance between samples: MSB = SSB/(£ — 1)
• Variance within samples: MSW = SSW/(n — k)
• Test statistic for a one-way ANOVA test:
F = MSB/MSW
Chapter 13 * Simple Linear Regression
• Simple linear regression model: y = A + Bx + e
• Estimated simple linear regression model: y = a + bx
Chapter 14 • Multiple Regression
Formulas for Chapter 14 along with the chapter are on the
Web site for the text.
Chapter 15 • Nonparametric Methods
Formulas for Chapter 15 along with the chapter are on the
Web site for the text.
Table IV Standard Normal Distribution Table
The entries in this table give the
cumulative area under the standard
normal curve to the left of z with the
values of z equal to or negative.
z
.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
-3.4
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0002
-3.3
.0005
.0005
.0005
.0004
.0004
.0004
.0004
.0004
.0004
.0003
-3.2
.0007
.0007
.0006
.0006
.0006
.0006
.0006
.0005
.0005
.0005
-3.1
.0010
.0009
.0009
.0009
.0008
.0008
.0008
.0008
.0007
.0007
-3.0
.0013
.0013
.0013
.0012
.0012
.0011
.0011
.0011
.0010
.0010
-2.9
.0019
.0018
.0018
.0017
.0016
.0016
.0015
.0015
.0014
.0014
-2.8
.0026
.0025
.0024
.0023
.0023
.0022
.0021
.0021
.0020
.0019
-2.7
.0035
.0034
.0033
.0032
.0031
.0030
.0029
.0028
.0027
.0026
-2.6
.0047
.0045
.0044
.0043
.0041
.0040
.0039
.0038
.0037
.0036
-2.5
.0062
.0060
.0059
.0057
.0055
.0054
.0052
.0051
.0049
.0048
-2.4
.0082
.0080
.0078
.0075
.0073
.0071
.0069
.0068
.0066
.0064
-2.3
.0107
.0104
.0102
.0099
.0096
.0094
.0091
.0089
.0087
.0084
-2.2
.0139
.0136
.0132
.0129
.0125
.0122
.0119
.0116
.0113
.0110
-2.1
.0179
.0174
.0170
.0166
.0162
.0158
.0154
.0150
.0146
.0143
-2.0
.0228
.0222
.0217
.0212
.0207
.0202
.0197
.0192
.0188
.0183
-1.9
.0287
.0281
.0274
.0268
.0262
.0256
.0250
.0244
.0239
.0233
-1.8
.0359
.0351
.0344
.0336
.0329
.0322
.0314
.0307
.0301
.0294
-1.7
.0446
.0436
.0427
.0418
.0409
.0401
.0392
.0384
.0375
.0367
-1.6
.0548
.0537
.0526
.0516
.0505
.0495
.0485
.0475
.0465
.0455
-1.5
.0668
.0655
.0643
.0630
.0618
.0606
.0594
.0582
.0571
.0559
-1.4
.0808
.0793
.0778
.0764
.0749
.0735
.0721
.0708
.0694
.0681
-1.3
.0968
.0951
.0934
.0918
.0901
.0885
.0869
.0853
.0838
.0823
-1.2
.1151
.1131
.1112
.1093
.1075
.1056
.1038
.1020
.1003
.0985
-1.1
.1357
.1335
.1314
.1292
.1271
.1251
.1230
.1210
.1190
.1170
-1.0
.1587
.1562
.1539
.1515
.1492
.1469
.1446
.1423
.1401
.1379
-0.9
.1841
.1814
.1788
.1762
.1736
.1711
.1685
.1660
.1635
.1611
-0.8
.2119
.2090
.2061
.2033
.2005
.1977
.1949
.1922
.1894
.1867
-0.7
.2420
.2389
.2358
.2327
.2296
.2266
.2236
.2206
.2177
.2148
-0.6
.2743
.2709
.2676
.2643
.2611
.2578
.2546
.2514
.2483
.2451
-0.5
.3085
.3050
.3015
.2981
.2946
.2912
.2877
.2843
.2810
.2776
-0.4
.3446
.3409
.3372
.3336
.3300
.3264
.3228
.3192
.3156
.3121
-0.3
.3821
.3783
.3745
.3707
.3669
.3632
.3594
.3557
.3520
.3483
-0.2
.4207
.4168
.4129
.4090
.4052
.4013
.3974
.3936
.3897
.3859
-0.1
.4602
.4562
.4522
.4483
.4443
.4404
.4364
.4325
.4286
.4247
0.0
.5000
.4960
.4920
.4880
.4840
.4801
.4761
.4721
.4681
.4641
(continued on next page)
Table IV Standard Normal Distribution Table (continued from previous page)
The entries in this table give the
cumulative area under the standard
normal curve to the left of z with the
values of z equal to or positive.
z
.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
0.0
.5000
.5040
.5080
.5120
.5160
.5199
.5239
.5279
.5319
.5359
1
5398
5438
5478
5517
5557
5596
5636
5675
.5714
5753
2
5793
.«j i y o
5832
5871
5910
5948
5987
6026
6064
6103
.6141
3
6179
.6217
6255
6?93
633 1
6368
6406
6443
6480
651 7
0.4
.6554
.6591
.6628
.6664
.6700
.6736
.6772
.6808
.6844
.6879
0.5
.6915
.6950
.6985
.7019
.7054
.7088
.7123
.7157
.7190
.7224
0.6
.7257
.7291
.7324
.7357
.7389
.7422
.7454
.7486
.7517
.7549
7
7580
.761 1
.7642
7673
.7704
.7734
.7764
7794
7823
785?
8
7881
791
■ /71U
7939
. / y oy
. i y\j i
7995
. / yy
80?3
8051
8078
8106
. O 1 \J\J
8 1 33
.Ol JJ
0.9
.8159
.8186
.8212
.8238
.8264
.8289
.8315
.8340
.8365
.8389
1.0
.8413
.8438
.8461
.8485
.8508
.8531
.8554
.8577
.8599
.8621
1.1
.8643
.8665
.8686
.8708
.8729
.8749
.8770
.8790
.8810
.8830
1.2
8849
8869
8888
8907
89? 5
8944
896?
8980
8997
. oyy i
901 5
1 3
903?
9049
9066
908?
9099
. y\jyy
9115
.711J
9131
9147
916?
. y 1 V.;z_
9177
.y i / /
1.4
.9192
.9207
.9222
.9236
.9251
.9265
.9279
.9292
.9306
.9319
1.5
.9332
.9345
.9357
.9370
.9382
.9394
.9406
.9418
.9429
.9441
1 f<
1 .u
Q/LS?
.yHOZ
.7tUJ
QA1A
QSOS
.yjuj
.yj i j
.7JZJ
.yjjj
.yjH-j
1 7
.7JOA
. yjy i
. y~>yy
.7UUO
. "O 1 u
Q69S
. 7UiJ
1 8
1 .0
.yojo
.yOOH
Q67 1
. yo / 1
.yo / o
.yooo
.7\J7J
.yoyy
.y / uo
1.9
.9713
.9719
.9726
.9732
.9738
.9744
.9750
.9756
.9761
.9767
2.0
.9772
.9778
.9783
.9788
.9793
.9798
.9803
.9808
.9812
.9817
Z. 1
Q89 1
.yozi
.yozo
.yoju
.yo
.yo jo
.yo^-z
.yo'+O
.yoju
QR^zL
.yoj^f
QR^7
.yoj /
9 9
z.z
.yooi
.yoo ,: f
.yooo
QR7 1
.yo / 1
.yo / J
.yo /o
.yoo 1
QRRzL
.yoo^-
QRR7
.yoo /
.yoyu
2.3
.9893
.9896
.9898
.9901
.9904
.9906
.9909
.9911
.9913
.9916
2.4
.9918
.9920
.9922
.9925
.9927
.9929
.9931
.9932
.9934
.9936
2.5
.9938
.9940
.9941
.9943
.9945
.9946
.9948
.9949
.9951
.9952
2.6
.9953
.9955
.9956
.9957
.9959
.9960
.9961
.9962
.9963
.9964
2.7
.9965
.9966
.9967
.9968
.9969
.9970
.9971
.9972
.9973
.9974
2.8
.9974
.9975
.9976
.9977
.9977
.9978
.9979
.9979
.9980
.9981
2.9
.9981
.9982
.9982
.9983
.9984
.9984
.9985
.9985
.9986
.9986
3.0
.9987
.9987
.9987
.9988
.9988
.9989
.9989
.9989
.9990
.9990
3.1
.9990
.9991
.9991
.9991
.9992
.9992
.9992
.9992
.9993
.9993
3.2
.9993
.9993
.9994
.9994
.9994
.9994
.9994
.9995
.9995
.9995
3.3
.9995
.9995
.9995
.9996
.9996
.9996
.9996
.9996
.9996
.9997
3.4
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9998
This is Table IV of Appendix C.
Table V The t Distribution Table
The entries in this table give the critical values
of t for the specified number of degrees
of freedom and areas in the right tail.
o t
Area in the Right Tail under the t Distribution Curve
df
.10
.05
.025
.01
.005
.001
1
3.078
6.314
12.706
31.821
63.657
318.309
2
1.886
2.920
4.303
6.965
9.925
22.327
3
1.638
2.353
3.182
4.541
5.841
10.215
4
1.533
2.132
2.776
3.747
4.604
7.173
5
1.476
2.015
2.571
3.365
4.032
5.893
6
1.440
1.943
2.447
3.143
3.707
5.208
7
1.415
1.895
2.365
2.998
3.499
4.785
8
1.397
1.860
2.306
2.896
3.355
4.501
9
1.383
1.833
2.262
2.821
3.250
4.297
10
1.372
1.812
2.228
2.764
3.169
4.144
11
1.363
1.796
2.201
2.718
3.106
4.025
12
1.356
1.782
2.179
2.681
3.055
3.930
13
1.350
1.771
2.160
2.650
3.012
3.852
14
1.345
1.761
2.145
2.624
2.977
3.787
15
1.341
1.753
2.131
2.602
2.947
3.733
16
1.337
1.746
2.120
2.583
2.921
3.686
17
1.333
1.740
2.110
2.567
2.898
3.646
18
1.330
1.734
2.101
2.552
2.878
3.610
19
1.328
1.729
2.093
2.539
2.861
3.579
20
1.325
1.725
2.086
2.528
2.845
3.552
21
1.323
1.721
2.080
2.518
2.831
3.527
22
1.321
1.717
2.074
2.508
2.819
3.505
23
1 319
1 714
1. / It
2 069
? 500
2 807
3 485
24
1.318
1.711
2.064
2.492
2.797
3.467
25
1.316
1.708
2.060
2.485
2.787
3.450
26
1.315
1.706
2.056
2.479
2.779
3.435
27
1.314
1.703
2.052
2.473
2.771
3.421
28
1.313
1.701
2.048
2.467
2.763
3.408
29
1.311
1.699
2.045
2.462
2.756
3.396
30
1.310
1.697
2.042
2.457
2.750
3.385
31
1.309
1.696
2.040
2.453
2.744
3.375
32
1.309
1.694
2.037
2.449
2.738
3.365
33
1.308
1.692
2.035
2.445
2.733
3.356
34
1.307
1.691
2.032
2.441
2.728
3.348
35
1.306
1.690
2.030
2.438
2.724
3.340
(continued on
next page)
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
00
The f Distribution Table (continued from previous page)
Area in the Right Tail under the t Distribution Curve
.10
OS
.025
.01
.005
.001
.306
1.688
2.028
2.434
2.719
3.333
.305
1.687
2.026
2.431
2.715
3.326
.304
1.686
2.024
2.429
2.712
3.319
o A/i
.304
1.685
O AO*}
2.023
O A 0£
2.426
O 7AO
2. /08
O O 1 o
3.313
.303
1 £ o A
1.684
O AO 1
2.021
O A OO
2.423
o nf\A
2. /04
O OAT
3.30/
.303
1.683
2.020
2.421
2.701
3.301
.302
1.682
2.018
2.418
2.698
3.296
.302
1.681
2.017
2.416
2.695
3.291
OA 1
.301
1 £ OA
1.680
O A 1 C
2.015
O A 1 A
Z.414
o £no
2.692
O O 0£
3.286
OA 1
.301
1.6 /9
O A 1 /I
2.014
O A 1 O
2.412
O £AA
2.690
O O 1
3.281
.300
1.679
2.013
2.410
2.687
3.277
.300
1.678
2.012
2.408
2.685
3.273
.299
1.677
2.011
2.407
2.682
3.269
OAA
.299
i fin
1.0/ /
O A 1 A
2.010
O A AC
2.405
O £ OA
2.680
0£ c
3.265
OAA
.299
1.6 lb
O AAA
2.009
O A AO
2.403
2.6/8
O O^ 1
3.261
.298
1.675
2.008
2.402
2.676
3.258
.298
1.675
2.007
2.400
2.674
3.255
.298
1.674
2.006
2.399
2.672
3.251
om
.29 /
1 en a
1.6 /4
O AAC
2.005
O OAT
2.39/
2.6/0
O /I o
3.248
om
.29/
1 ^to
1.6/3
O A A A
2.004
O A£
2.396
O o
2.668
3.245
.297
1.673
2.003
2.395
2.667
3.242
.297
1.672
2.002
2.394
2.665
3.239
.296
1.672
2.002
2.392
2.663
3.237
OA£
.296
1.6/1
O AA 1
2.001
O A 1
2.391
O ££0
2.662
O O O /I
3.234
.296
1.6/1
O AAA
2.000
O OAA
2.390
O ££A
2.660
o oo o
5.151
.296
1.670
2.000
2.389
2.659
3.229
.295
1.670
1.999
2.388
2.657
3.227
.295
1.669
1.998
2.387
2.656
3.225
one
.295
1 ££A
1.669
1 AAO
1.998
O O O^
2.386
2.655
o ooo
3.223
one
.295
1 ££A
1.669
1 AA*7
1.99/
o o o c
2.385
O C A
2.654
O OO A
3.220
.295
1.668
1.997
2.384
2.652
3.218
.294
1.668
1.996
2.383
2.651
3.216
.294
1.668
1.995
2.382
2.650
3.214
.294
1.667
1.995
2.382
2.649
3.213
.294
1.667
1.994
2.381
2.648
3.211
.294
1.667
1.994
2.380
2.647
3.209
.293
1.666
1.993
2.379
2.646
3.207
.293
1.666
1.993
2.379
2.645
3.206
.293
1.666
1.993
2.378
2.644
3.204
.293
1.665
1.992
2.377
2.643
3.202
.282
1.645
1.960
2.326
2.576
3.090
Table V of Appendix C.
PLUS
www.wileyplus.com
This online teaching and learning environment
integrates the entire digital textbook with the
most effective instructor and student resources
to fit every learning style.
r
With WileyPLUS: ^
D Students achieve concept
mastery in a rich,
structured environment
that's available 24/7
3
Instructors personalize and manage
their course more effectively with
assessment, assignments, grade
tracking, and more
manage time bette
• study smarter
• save money
J
r
From multiple study paths, to self-assessment, to a wealth of interactive
visual and audio resources, WileyPLUS gives you everything you need to
personalize the teaching and learning experience.
>Find out how to MAKE IT YOURS»
www.wileyplus.com
WILEY
PLUS
p ALL THE HELP, RESOURCES, AND PERSONAL SUPPORT
YOU AND YOUR STUDENTS NEED!
a ct DAY OF
■ CLASS
... AND BEYOND!
2-Minute Tutorials and all
of the resources you & your
students need to get started
www.wileyplus.com/firstday
WILEY
PLUS
QuickStart
Pre-loaded, ready-to-use
assignments and presentations
www.wiley.com/college/quickstart
WILEY
WILEY
PLUS
Student Partner Program
Student support from an
experienced student user
Ask your local representative
for details!
Collaborate with your colleagues,
find a mentor, attend virtual and live
events, and view resources
www.WhereFacultyConnect.com
i
V
Technical Support 24/7 Your WileyPLUS
FAQs, online chat, Account Manager
and phone support Training and implementation support
www.wileyplus.com/support www.wileyplus.com/accountmanager
PLUS MAKE IT YOURS!
www.wileyplus.com
Seventh Edition
INTRODUCTORY STATISTICS
Seventh Edition
INTRODUCTORY STATISTICS
PREM S. MANN
EASTERN CONNECTICUT STATE UNIVERSITY
WITH THE HELP OF
CHRISTOPHER JAY LACKE
ROWAN UNIVERSITY
WILEY
JOHN WILEY & SONS, INC.
Vice President & Executive Publisher Laurie Rosatone
Project Editors Jenn Albanese and Ellen Keohane
Production Manager Dorothy Sinclair
Senior Production Editor Valerie A. Vargas
Marketing Manager Sarah Davis
Creative Director Harry Nolan
Designer Director Jeof Vita
Production Management Services Aptara®, Inc.
Senior Illustration Editor Anna Melhorn
Photo Associate Sarah Wilkin
Editorial Assistant Beth Pearson
Media Editors Melissa Edwards and Ari Wolfe
Cover Photo Credit © James Leynse/©Corbis
This book was set in 10/12 Times Roman by Aptara®, Inc. and printed and bound
by Courier-Kendallville. The cover was printed by Courier-Kendallville.
This book is printed on acid free paper. °°
Copyright © 2010, 2007, 2004, 2001, John Wiley & Sons, Inc. All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any
means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as permitted under Sections
107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the
Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc.
222 Rosewood Drive, Danvers, MA 01923, website www.copyright.com. Requests to the Publisher for
permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., Ill River Street,
Hoboken, NJ 07030-5774, (201)748-6011, fax (201)748-6008, website http://www.wiley.com/go/permissions.
Evaluation copies are provided to qualified academics and professionals for review purposes only, for use
in their courses during the next academic year. These copies are licensed and may not be sold or transferred to
a third party. Upon completion of the review period, please return the evaluation copy to Wiley. Return instruc-
tions and a free of charge return shipping label are available at www.wiley.com/go/returnlabel. Outside of the
United States, please contact your local representative.
Library of Congress Cataloging in Publication Data
ISBN- 13 978-0-470-44466-5 (cloth)
ISBN- 13 978-0-470-55663-4 (Binder Ready Version )
Printed in the United States of America
10 987654 3 21
To my mother
and
to the memory of my father
BREEACE
Introductory Statistics is written for a one- or two-semester first course in applied statistics.
This book is intended for students who do not have a strong background in mathematics. The
only prerequisite for this text is knowledge of elementary algebra.
Today, college students from almost all fields of study are required to take at least one
course in statistics. Consequently, the study of statistical methods has taken on a prominent
role in the education of students from a variety of backgrounds and academic pursuits. From
the first edition, the goal of Introductory Statistics has been to make the subject of statistics
interesting and accessible to a wide and varied audience. Three major elements of this text
support this goal:
1. Realistic content of its examples and exercises, drawing from a comprehensive range of ap-
plications from all facets of life
2. Clarity and brevity of presentation
3. Soundness of pedagogical approach
These elements are developed through the interplay of a variety of significant text features.
The feedback received from the users of the sixth edition of Introductory Statistics has
been very supportive and encouraging. Positive experiences reported by instructors and stu-
dents have served as evidence that this text offers an interesting and accessible approach to
statistics — the author's goal from the very first edition. The author has pursued the same goal
through the refinements and updates in this seventh edition, so that Introductory Statistics
can continue to provide a successful experience in statistics to a growing number of students
and instructors.
New to the Seventh Edition
The following are some of the changes made in the seventh edition:
■ A large number of the examples and exercises are new, providing new and varied ways for
students to practice statistical concepts.
■ Most of the case studies are new or revised, drawing on current uses of statistics in areas
of student interest.
■ New chapter opening images and questions incorporate real data in familiar situations.
■ New data are integrated throughout, reinforcing the vibrancy of statistics and the relevance
of statistics to student lives right now.
■ The Technology Instruction sections have been heavily revised to support the use of the lat-
est versions of the TI-84/84+, Minitab, and Excel.
■ Many of the Uses and Misuses sections are either new or have been updated.
■ Many of the Decide for Yourself sections are either new or have been updated.
■ Several new Miniprojects have been added.
■ A large number of new Technology Assignments have been added.
vii
VIII
Preface
Style and Pedagogy
Hallmark Features of this Text
Clear and Concise Exposition The explanation of statistical methods and concepts is clear and
concise. Moreover, the style is user-friendly and easy to understand. In chapter introductions and
in transitions from section to section, new ideas are related to those discussed earlier.
Thorough Examples
Examples The text contains a wealth of examples, more than 200 in 15 chapters and Appendix
A. The examples are usually presented in a format showing a problem and its solution. They
are well sequenced and thorough, displaying all facets of concepts. Furthermore, the examples
capture students' interest because they cover a wide variety of relevant topics. They are based
on situations that practicing statisticians encounter every day. Finally, a large number of exam-
ples are based on real data taken from sources such as books, government and private data
sources and reports, magazines, newspapers, and professional journals.
Step-by-Step Solutions
Solutions A clear, concise solution follows each problem presented in an example. When the
solution to an example involves many steps, it is presented in a step-by-step format. For in-
stance, examples related to tests of hypothesis contain five steps that are consistently used to
solve such examples in all chapters. Thus, procedures are presented in the concrete settings of
applications rather than as isolated abstractions. Frequently, solutions contain highlighted re-
marks that recall and reinforce ideas critical to the solution of the problem. Such remarks add
to the clarity of presentation.
Enlightening Pedagogy
Margin Notes for Examples A margin note appears beside each example that briefly de-
scribes what is being done in that example. Students can use these margin notes to assist them
as they read through sections and to quickly locate appropriate model problems as they work
through exercises.
Frequent Use of Diagrams Concepts can often be made more understandable by describing
them visually with the help of diagrams. This text uses diagrams frequently to help students
understand concepts and solve problems. For example, tree diagrams are used extensively in
Chapters 4 and 5 to assist in explaining probability concepts and in computing probabilities.
Similarly, solutions to all examples about tests of hypothesis contain diagrams showing rejec-
tion regions, nonrejection regions, and critical values.
Highlighting Definitions of important terms, formulas, and key concepts are enclosed in
colored boxes so that students can easily locate them.
Cautions Certain items need special attention. These may deal with potential trouble spots
that commonly cause errors, or they may deal with ideas that students often overlook. Special
emphasis is placed on such items through the headings Remember, An Observation, or Warn-
ing. An icon is used to identify such items.
Realistic Applications
Case Studies Case studies, which appear in almost all chapters, provide additional illustra-
tions of the applications of statistics in research and statistical analysis. Most of these case stud-
ies are based on articles/snapshots published in journals, magazines, or newspapers. All case
studies are based on real data.
Abundant Exercises Exercises and Supplementary Exercises The text contains an abundance of exercises
(excluding Technology Assignments) — approximately 1500 in 15 chapters and Appendix A.
Moreover, a large number of these exercises contain several parts. Exercise sets appearing at the
end of each section (or sometimes at the end of two or three sections) include problems on the
topics of that section. These exercises are divided into two parts: Concepts and Procedures that
emphasize key ideas and techniques, and Applications that use these ideas and techniques in
concrete settings. Supplementary exercises appear at the end of each chapter and contain exer-
cises on all sections and topics discussed in that chapter. A large number of these exercises are
based on real data taken from varied data sources such as books, government and private data
Preface IX
sources and reports, magazines, newspapers, and professional journals. Not merely do the exer-
cises given in the text provide practice for students, but the real data contained in the exercises
provide interesting information and insight into economic, political, social, psychological, and
other aspects of life. The exercise sets also contain many problems that demand critical thinking
skills. The answers to selected odd-numbered exercises appear in the Answers section at the back
of the book. Optional exercises are indicated by an asterisk (*).
Advanced Exercises All chapters (except Chapters 1 and 14) have a set of exercises that are Challenging Problems
of greater difficulty. Such exercises appear under the heading Advanced Exercises as part of the
Supplementary Exercises.
Uses and Misuses This feature at the end of each chapter (before the Glossary) points out Misconceptions and Pitfalls
common misconceptions and pitfalls students will encounter in their study of statistics and in
everyday life. Subjects highlighted include such diverse topics as the use of the word average
and grading on a curve.
Decide for Yourself This feature appears at the end of each chapter (except Chapter 1) just Open-ended Problems
before the Technology Instruction section. In this section, a real-world problem is discussed,
and questions are raised about this problem that readers are required to answer.
Glossary Each chapter has a glossary that lists the key terms introduced in that chapter, along Summary and Review
with a brief explanation of each term. Almost all the terms that appear in boldface type in the
text are in the glossary.
Self-Review Tests Each chapter contains a Self-Review Test, which appears immediately after Testing Yourself
the Supplementary Exercises. These problems can help students test their grasp of the concepts
and skills presented in respective chapters and monitor their understanding of statistical meth-
ods. The problems marked by an asterisk (*) in the Self-Review Tests are optional. The answers
to almost all problems of the Self-Review Tests appear in the Answer section.
Formula Card A formula card that contains key formulas from all chapters and the normal Key Formulas
distribution and t distribution tables is included at the beginning of the book.
Technology Usage At the end of each chapter is a section covering uses of three major tech- Technology Usage
nologies of statistics and probability: the TI-84, Minitab, and Excel. For each technology, stu-
dents are guided through performing statistical analyses in a step-by-step fashion, showing them
how to enter, revise, format, and save data in a spreadsheet, workbook, or named and un-
named lists, depending on the technology used. Illustrations and screen shots demonstrate the
use of these technologies. Additional detailed technology instruction is provided in the technol-
ogy manuals that are online at www.wiley.com/college/mann.
Technology Assignments Each chapter contains a few technology assignments that appear at Technology Assignments
the end of the chapter. These assignments can be completed using any of the statistical software.
Miniprojects Each chapter contains a few Miniprojects that appear just before the Decide it Miniprojects
Yourself sections. These Miniprojects are like very comprehensive exercises or ask students to
perform their own surveys and experiments. They provide practical applications of statistical
concepts to real life.
Data Sets A large number of data sets appear on the Web site for the text that is located Data Sets
at www.wiley.com/college/mann. These data sets include the data for various exercises in
the text and eight large data sets. These eight large data sets are collected from various
sources and they contain information on several variables. Many exercises and assignments
in the text are based on these data sets. These large data sets can also be used for instruc-
tor-driven analyses using a wide variety of statistical software packages as well as the TI-
84. These data sets are available on the Web site of the text in a variety of formats in-
cluding Minitab 1 , Excel, and text formats.
'Minitab is a registered trademark of Minitab, Inc., Quality Plaza, 1829 Pine Hall Road, State College, PA 16801-3008.
Phone: 814-238-3280.
X
Preface
Statistical Animations Statistical Animations In relevant places throughout the text, an icon alerts students to the
availability of a statistical animation. These animations illustrate statistical concepts in the text,
and can be found on the companion Web-Site.
GAISE Report Recommendations Adopted
In 2003, the American Statistical Association (ASA) funded the Guidelines for Assessment and
Instruction in Statistics Education (GAISE) Project to develop ASA-endorsed guidelines for as-
sessment and instruction in statistics for the introductory college statistics course. The report,
which can be found at www.amstat.org/education/gaise, resulted in the following series of rec-
ommendations for the first course in statistics and data analysis.
1. Emphasize statistical literacy and develop statistical thinking.
2. Use real data.
3. Stress conceptual understanding rather than mere knowledge of procedures.
4. Foster active learning in the classroom.
5. Use technology for developing concepts and analyzing data.
6. Use assessments to improve and evaluate student learning.
Here are a few examples of how this Introductory Statistics text can assist in helping you, the
instructor, in meeting the GAISE recommendations.
1. Many of the newer exercises require interpretation, not just a number. Graphical and nu-
meric summaries are combined in some new exercises in order to emphasize looking at the
whole picture, as opposed to using just one graph or one summary statistic.
2. The Decide for Yourself and Uses and Misuses features help to develop statistical thinking
and conceptual understanding.
3. All of the data sets in the exercises and in Appendix B are available on the book's Web site.
They have been formatted for a variety of statistical software packages. This eliminates the
need to enter data into the software. A variety of software instruction manuals also allows the
instructor to spend more time on concepts, and less time teaching how to use technology.
4. The Miniprojects help students to generate their own data by performing an experiment
and/or taking random samples from the large datasets given in Appendix B.
We highly recommend that all statistics instructors take the time to read the GAISE report. There
is a wealth of information in this report that can be used by everyone.
Web Site
http://www.wiley.com/college/mann
The Web site for this text provides additional resources for instructors and students. The fol-
lowing items are available on this Web-site:
• Formula Card
• Statistical Animations
• Computerized Test Bank
• Instructor's Solutions Manual
• PowerPoint Slides
• Data Sets (see Appendix B for a complete list of these data sets)
• Technology Resource Manuals.
• TI Graphing Calculator Manual
• Minitab Manual
• Excel Manual
Preface XI
These manuals provide step-by-step instructions, screen captures, and examples for using tech-
nology in the introductory statistics course. Also provided are exercise tables and indications of
which exercises from the text best lend themselves to the use of the package presented.
• Chapter 14: Multiple Regression
• Chapter 15: Nonparametric Methods
Using WileyPLUS
This online teaching and learning environment integrates the entire digital textbook with
the most effective instructor and student resources to fit every learning style. With Wiley-
PLUS:
• Students achieve concept mastery in a rich, structured environment that is available 24/7.
• Instructors personalize and manage their course more effectively with assessment, assign-
ments, grade tracking, and more.
WileyPLUS can complement the current textbook or replace the printed text altogether.
For Students
Personalize the learning experience:
Different learning styles, different levels of proficiency, different levels of preparation — each
of your students is unique. WileyPLUS empowers them to take advantage of their individual
strengths:
• Students receive timely access to resources that address their demonstrated needs and get
immediate feedback and remediation when needed.
• Integrated, multimedia resources provide multiple studypaths to fit each student's learning
preferences and encourage more-active learning.
• WileyPLUS includes many opportunities for self-assessment linked to the relevant portions
of the text. Students can take control of their learning and practice until they master the
material.
For Instructors
Personalize the teaching experience:
WileyPLUS empowers you with the tools and resources you need to make your teaching even
more effective:
• You can customize your classroom presentation with a wealth of resources and functional-
ity, from PowerPoint slides to a database of rich visuals. You can even add your own mate-
rials to your WileyPLUS course.
• With WileyPLUS you can identify those students who are falling behind and intervene ac-
cordingly, without having to wait for them to come to your office hours.
• WileyPLUS simplifies and automates such tasks as student performance assessment, making
assignments, scoring student work, keeping grades, and more.
Supplements
The following supplements are available to accompany this text:
■ Instructor's Solutions Manual (ISBN 978-0-470-57241-2) This manual contains com-
pete solutions to all of the exercises in the text.
■ Printed Test Bank (ISBN 978-0-470-57242-9) The printed copy of the test bank con-
tains a large number of multiple-choice questions, essay questions, and quantitative prob-
lems for each chapter.
Xii Preface
■ Computerized Test Bank All of the questions in the Printed Test Bank are available elec-
tronically and can be obtained from the publisher.
■ Student Solutions Manual (ISBN 978-0-470-57239-9) This manual contains complete
solutions to all of the odd-numbered exercises in the text.
■ Student Study Guide (ISBN 978-0-470-57240-5) This guide contains review material
for a first course in statistics. Special attention is given to the critical material for each chap-
ter. Reviews of mathematical notation and formulas are also included.
Acknowledgments
I thank the following reviewers of this and/or previous editions of this book, whose comments
and suggestions were invaluable in improving the text.
Alfred A. Akinsete
Marshall University
Scott S. Albert
College of DuPage
Michael R. Allen
Tennessee Technological University
Peter Arvanites
Rockland Community College
K. S. Asal
Broward Community College
Louise Audette
Manchester Community College
Nicole Betsinger
Arapahoe Community College
Joan Bookbinder
Johnson & Wales University
Dean Burbank
Gulf Coast Community College
Helen Burn
Highline Community College
Gerald Busald
San Antonio College
Peter A. Carlson
Delta College
Jayanta Chandra
University of Notre Dame
C. K. Chauhan
Indiana-Purdue University at Fort Wayne
James Curl
Modesto Community College
Gregory Daubenmire
Las Positas Community College
Joe DeMaio
Kennesaw State University
Fred H. Dorner
Trinity University, San Antonio
William D. Ergle
Roanoke College, Salem, Virginia
Ruby Evans
Santa Fe Community College
Ronald Ferguson
San Antonio College
James C. Ford
Anda Gadidov
Kennesaw State University
Frank Goulard
Portland Community College
Robert Graham
Jacksonville State University,
Jacksonville, Alabama
Larry Griffey
Florida Community College,
Jacksonville
Arjun K. Gupta
Bowling Green State University
David Gurney
Southeastern Louisiana University
Daesung Ha
Marshall University
A. Eugene Hileman
Northeastern State University,
Tahlequah, Oklahoma
John G. Horner
Cabrillo College
Virginia Horner
Diablo Valley College
Ina Parks S. Howell
Florida International University
John Haussermann
Monterey Peninsular College
Shana Irwin
University of North Texas
Gary S. Itzkowitz
Rowan State College
Joanna Jeneralczuk
University of Massachusetts, Amherst
Jean Johnson
Governors State University
Michael Karelius
American River College, Sacramento
Preface Xlii
Dix J. Kelly
Central Connecticut State University
Jong Sung Kim
Portland State University
Linda Kohl
University of Michigan, Ann Arbor
Martin Kotler
Pace University, Pleasantville, New York
Marlene Kovaly
Florida Community College, Jacksonville
Hillel Kumin
University of Oklahoma
Carlos de la Lama
San Diego City College
Rita Lindsay
Indian River State College
Gaurab Mahapatra
University of Akron
Richard McGowan
University of Scranton
Daniel S. Miller
Central Connecticut State University
Dorothy Miners
Brock University
Satya N. Mishra
University of South Alabama
Jeffrey Mock
Diablo Valley College
Luis Moreno
Broome Community College, Binghamton
Robert A. Nagy
University of Wisconsin, Green Bay
Sharon Navard
The College of New Jersey
Nhu T. Nguyen
New Mexico State University
Paul T. Nkansah
Florida Agricultural and Mechanical
University
Joyce Oster
Johnson and Wales University
Lindsay Packer
College of Charleston
Mary Parker
Austin Community College
Roger Peck
University of Rhode Island, Kingston
Chester Piascik
Bryant College, Smithfield
Joseph Pigeon
Villanova University
Cristina Popescue,
Grant MacEwan College
Aaron Robertson
Colgate University
Gerald Rogers
New Mexico State University, Las Cruces
Emily Ross
University of Missouri, St. Louis
Juana Sanchez
UCLA
Brunilda Santiago
Indian River State College
Phillis Schumacher
Bryant College, Smithfield
Kathryn Schwartz
Scottsdale Community College
Ronald Schwartz
Wilkes University, Wilkes-Barre
David Stark
University of Akron
Larry Stephens
University of Nebraska, Omaha
Bruce Trumbo
California State University, Hayward
Vasant Waikar
Miami University
Jean Weber
University of Arizona, Tucson
Terry Wilson
San Jacinto College, Pasadena
James Wright
Bucknell University
K. Paul Yoon
Fairleigh Dickinson University, Madison
Zhiyi Zhang
University of North Carolina
I express my thanks to the following for their contributions to earlier editions of this book
that made it better in many ways: Maryanne Clifford (Eastern Connecticut State University),
Gerald Geissert, Daniel S. Miller (Central Connecticut State University), and David Santana-
Ortiz (Rand Organization).
I extend my special thanks to Christopher Lacke of Rowan University, who contributed to
this edition in many significant ways. Without his help, this book would not be in this form. I take
this opportunity to thank Ann Ostberg for preparing the answers section for the back of the book
and for checking the text examples for math accuracy, and to Sandra Zirkes for checking the an-
swers section for accuracy. In addition, I thank Eastern Connecticut State University for all the
support I received.
XiV Preface
It is of utmost importance that a textbook be accompanied by complete and accurate
supplements. I take pride in mentioning that the supplements prepared for this text possess
these qualities and much more. I thank the authors of all these supplements.
It is my pleasure to thank all the professionals at John Wiley with whom I enjoyed work-
ing during this revision. Among them are Laurie Rosatone (Vice President and Executive
Publisher), Jackie Henry (Full Service Manager), Jeof Vita (Art Director), Sarah Wilkin
(Photo Associate), Dorothy Sinclair (Production Manager), Valerie Vargas (Senior Production
Editor), Ellen Keohane (Editor), Beth Pearson (Editorial Assistant), Melissa Edwards (Media
Editor), Ari Wolfe (Media Project Editor), and Sarah Davis (Marketing Manager), Harry
Nolan (Creative Director), Anna Melhorn (Senior Illustration Editor). Lastly but most
importantly I extend my most heartfelt thanks to Jenn Albanese (Project Editor) whose support
and guidance was of immense help during this revision.
Any suggestions from readers for future revisions would be greatly appreciated. Such
suggestions can be sent to the author at mann@easternct.edu or premmann@yahoo.com.
Prem S. Mann
Willimantic, CT
November 2009
CONTENTS
Chapter 1 Introduction 1
1.1 What is Statistics? 2
1 .2 Types of Statistics 2
Case Study 1-1 2008 U.S. Patent Leaders 3
Case Study 1-2 TV Commercials and Holiday Shopping 4
1 .3 Population Versus Sample 5
Case Study 1 -3 On Road, It's "Do As I Say, Not As I Do" 7
1 .4 Basic Terms 8
1.5 Types of Variables 10
1.6 Cross-Section Versus Time-Series Data 13
1.7 Sources of Data 14
1.8 Summation Notation 15
Uses and Misuses/Glossary/Supplementary Exercises/Self-Review Test/Mini-Project/
Technology Instruction/Technology Assignments
Chapter 2 Organizing and Graphing Data 27
2.1 Raw Data 28
2.2 Organizing and Graphing Qualitative Data 28
Case Study 2-1 Career Choices for High School Students 31
Case Study 2-2 In or Out in 30 Minutes 32
2.3 Organizing and Graphing Quantitative Data 35
Case Study 2-3 Morning Grooming 40
2.4 Shapes of Histograms 44
2.5 Cumulative Frequency Distributions 51
2.6 Stem-and-Leaf Displays 54
2.7 Dotplots 58
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/
Mini-Projects/Decide for Yourself/Technology Instruction/Technology Assignments
XV
XVi Contents
Chapter 3 Numerical Descriptive Measures 79
3.1 Measures of Central Tendency for Ungrouped Data 80
Case Study 3-1 Average Attendance at Baseball Games 83
Case Study 3-2 The Gender Pay Gap 85
3.2 Measures of Dispersion for Ungrouped Data 92
3.3 Mean, Variance, and Standard Deviation for Grouped Data 98
3.4 Use of Standard Deviation 105
Case Study 3-3 Here Comes the SD 108
3.5 Measures of Position 110
3.6 Box-and-Whisker Plot 115
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Appendix 3.1 /Self-Review Test/
Mini-Projects/Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 4 Probability 137
4.1 Experiment, Outcomes, and Sample Space 138
4.2 Calculating Probability 143
4.3 Counting Rule 149
4.4 Marginal and Conditional Probabilities 150
Case Study 4-1 Rolling Stops 153
4.5 Mutually Exclusive Events 154
4.6 Independent Versus Dependent Events 155
4.7 Complementary Events 156
4.8 Intersection of Events and the Multiplication Rule 161
Case Study 4-2 Baseball Players have "Slumps" and "Streaks" 167
4.9 Union of Events and the Addition Rule 171
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 5 Discrete Random Variables and Their Probability
Distributions 191
5.1 Random Variables 192
5.2 Probability Distribution of a Discrete Random Variable 194
5.3 Mean of a Discrete Random Variable 201
5.4 Standard Deviation of a Discrete Random Variable 202
Case Study 5-1 Aces High Instant Lottery Game-20th Edition 203
5.5 Factorials, Combinations, and Permutations 208
Case Study 5-2 Playing Lotto 212
5.6 The Binomial Probability Distribution 214
5.7 The Hypergeometric Probability Distribution 226
5.8 The Poisson Probability Distribution 230
Case Study 5-3 Ask Mr. Statistics 233
Case Study 5-4 Living and Dying in the USA 235
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Contents XVI i
Chapter 6 Continuous Random Variables and the Normal Distribution 250
6.1 Continuous Probability Distribution 251
6.2 The Normal Distribution 254
Case Study 6-1 Distribution of Time Taken to Run a Road Race 255
6.3 The Standard Normal Distribution 259
6.4 Standardizing a Normal Distribution 267
6.5 Applications of the Normal Distribution 273
6.6 Determining the z and x Values When an Area Under the Normal Distribution Curve
is Known 278
6.7 The Normal Approximation to the Binomial Distribution 283
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 7 Sampling Distributions 300
7.1 Population and Sampling Distributions 301
7.2 Sampling and Nonsampling Errors 303
7.3 Mean and Standard Deviation of x 306
7.4 Shape of the Sampling Distribution of x 310
7.5 Applications of the Sampling Distribution of x 316
7.6 Population and Sample Proportions 321
7.7 Mean, Standard Deviation, and Shape of the Sampling Distribution of p 323
7.8 Applications of the Sampling Distribution of p 328
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 8 Estimation of the Mean and Proportion 340
8.1 Estimation: An Introduction 341
8.2 Point and Interval Estimates 342
8.3 Estimation of a Population Mean: a Known 344
Case Study 8-1 Raising a Child 349
8.4 Estimation of a Population Mean: a Not Known 354
8.5 Estimation of a Population Proportion: Large Samples 362
Case Study 8-2 Which Sound Is the Most Frustrating to Hear? 365
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 9 Hypothesis Tests About the Mean and Proportion 381
9.1 Hypothesis Tests: An Introduction 382
9.2 Hypothesis Tests About fi: a Known 390
Case Study 9-1 How Crashes Affect Auto Premiums 399
9.3 Hypothesis Tests About fi: a Not Known 404
9.4 Hypothesis Tests About a Population Proportion: Large Samples 414
XViii Contents
Case Study 9-2 Favorite Seat in the Plane 420
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 1 Estimation and Hypothesis Testing: Two Populations 439
10.1 Inferences About the Difference Between Two Population Means for Independent Samples:
cr, and a 2 Known 440
10.2 Inferences About the Difference Between Two Population Means for Independent Samples:
cr, and cr 2 Unknown but Equal 447
Case Study 10-1 Average Compensation for Accountants 454
10.3 Inferences About the Difference Between Two Population Means for Independent Samples:
cr, and cr 2 Unknown and Unequal 457
10.4 Inferences About the Difference Between Two Population Means for Paired Samples 464
10.5 Inferences About the Difference Between Two Population Proportions for Large and Independent
Samples 473
Case Study 10-2 Is Vacation Important? 478
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 1 Chi-Square Tests 498
11.1 The Chi-Square Distribution 499
1 1 .2 A Goodness-of-Fit Test 502
Case Study 11-1 What Is Your Favorite Season? 508
11.3 Contingency Tables 511
11.4 A Test of Independence or Homogeneity 511
1 1 .5 Inferences About the Population Variance 523
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 1 2 Analysis of Variance 541
12.1 The F Distribution 542
12.2 One-Way Analysis of Variance 544
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 1 3 Simple Linear Regression 564
13.1 Simple Linear Regression Model 565
13.2 Simple Linear Regression Analysis 567
Case Study 13-1 Regression of Heights and Weights of NBA Players 574
13.3 Standard Deviation of Random Errors 581
13.4 Coefficient of Determination 582
13.5 Inferences About B 587
13.6 Linear Correlation 592
Contents Xix
13.7 Regression Analysis: A Complete Example 599
13.8 Using the Regression Model 606
13.9 Cautions in Using Regression 609
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Chapter 1 4 Multiple Regression
This chapter is not included in this text but is available for download on the Web site at www.wiley.com/college/mann.
14.1 Multiple Regression Analysis
14.2 Assumptions of the Multiple Regression Model
14.3 Standard Deviation of Errors
14.4 Coefficient of Multiple Determination
14.5 Computer Solutions of Multiple Regression
Uses and Misuses/Glossary/Self-Review Test/Mini-Projects/Decide for Yourself
Chapter 1 5 Nonparametric Methods
This chapter is not included in this text but is available for download on the Web site at www.wiley.com/college/mann.
15.1 The Sign Test
15.2 The Wilcoxon Signed-Rank Test for Two Dependent Samples
15.3 The Wilcoxon Rank Sum Test for Two Independent Samples
15.4 The Kruskal-Wallis Test
15.5 The Spearman Rho Rank Correlation Coefficient Test
15.6 The Runs Test for Randomness
Uses and Misuses/Glossary/Supplementary Exercises/Advanced Exercises/Self-Review Test/Mini-Projects/
Decide for Yourself/Technology Instruction/Technology Assignments
Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments Al
A.1 Sources of Data Al
A.l.l Internal Sources Al
A.1.2 External Sources A2
A.1.3 Surveys and Experiments A2
Case Study A-l Is It a Simple Question? A3
A.2 Sample Surveys and Sampling Techniques A4
A.2.1 Why Sample? A4
A.2.2 Random and Nonrandom Samples A4
A.2.3 Sampling and Nonsampling Errors A5
A.2.4 Random Sampling Techniques A8
A.3 Design of Experiments A9
Case Study A-2 Do Antibacterial Soaps Work? A13
Exercises / Advanced Exercises / Glossary
Appendix B Explanation of Data Sets Bl
Data Set I: City Data Bl
Data Set II: Data on States B2
XX Contents
Data Set III: NBA Data B2
Data Set IV: Manchester (Connecticut) Road Race Data B3
Data Set V: Sample of 500 Observations Selected from Data Set IV B3
Data Set VI: Data on Movies B3
Data Set VII: Standard & Poor's 100 Index Data B4
Data Set VIII: McDonald's Data B4
Appendix C Statistical Tables CI
I Table of Binomial Probabilities C2
II Values of e A Cll
III Table of Poisson Probabilities CI 3
IV Standard Normal Distribution Table CI 9
V The f Distribution Table C21
VI Chi-Square Distribution Table C23
VII The F Distribution Table C24
Tables VIII through XII along with Chapters 14 and 15 are available on the Web site of this text.
VIII Critical Values of X for the Sign Test
IX Critical Values of T for the Wilcoxon Signed-Rank Test
X Critical Values of T for the Wilcoxon Rank Sum Test
XI Critical Values for the Spearman Rho Rank Correlation Coefficient Test
XII Critical Values for a Two-Tailed Runs Test with a = .05
Answers to Selected Odd-Numbered Exercises and Self-Review Tests AN1
Photo Credits PCI
Index II
Seventh Edition
INTRODUCTORY STATISTICS
Chapter
Introduction
Do you feel compelled to shop at specific stores during the holidays? If yes, do you know why?
Do you think TV commercials have anything to do with it? In a survey conducted by the National
Retail Federation, only 18% of adults said that TV commercials influence them to shop at specific
stores. Do you find this value to be surprising? (See Case Study 1-2).
The study of statistics has become more popular than ever over the past four decades or so. The in-
creasing availability of computers and statistical software packages has enlarged the role of statistics
as a tool for empirical research. As a result, statistics is used for research in almost all professions,
from medicine to sports. Today, college students in almost all disciplines are required to take at least
one statistics course. Almost all newspapers and magazines these days contain graphs and stories on
statistical studies. After you finish reading this book, it should be much easier to understand these
graphs and stories.
Every field of study has its own terminology. Statistics is no exception. This introductory chapter
explains the basic terms of statistics. These terms will bridge our understanding of the concepts and
techniques presented in subsequent chapters.
1.1 What Is Statistics?
1 .2 Types of Statistics
Case Study 1-1 2008 U.S.
Patent Leaders
Case Study 1-2 TV Commercials
and Holiday Shopping
1.3 Population Versus Sample
Case Study 1-3 On Road, It's
"Do as I Say, Not as I Do"
1 .4 Basic Terms
1 .5 Types of Variables
1 .6 Cross-Section Versus
Time-Series Data
1.7 Sources of Data
1.8 Summation Notation
1
2 Chapter 1 Introduction
1.1 What Is Statistics?
The word statistics has two meanings. In the more common usage, statistics refers to numeri-
cal facts. The numbers that represent the income of a family, the age of a student, the per-
centage of passes completed by the quarterback of a football team, and the starting salary of a
typical college graduate are examples of statistics in this sense of the word. A 1988 article in
U.S. News & World Report declared "Statistics are an American obsession." 1 During the 1988
baseball World Series between the Los Angeles Dodgers and the Oakland A's, the then NBC
commentator Joe Garagiola reported to the viewers numerical facts about the players' per-
formances. In response, fellow commentator Vin Scully said, "I love it when you talk statis-
tics." In these examples, the word statistics refers to numbers.
The following examples present some statistics:
1. During the 43rd Super Bowl on February 1, 2009, NBC charged $3 million for a 30-second
commercial.
2. New York City mayor, Michael Bloomberg, gave $239 million to charity in 2008.
3. According to the Chronicle of Philanthropy, Wal-Mart gave $337.9 million to charity in
2007.
4. According to the U.S. Department of Agriculture, about 900 million roses were imported
from Colombia to the United States in 2005.
5. According to a 2008 SHRM Employee Benefit Survey, 3% of large companies allow pets
at work.
6. According to the Centers for Disease Control and Prevention, flu costs the United States
about $87 billion a year in terms of direct medical costs, loss of life, and reduced quality
of life.
The second meaning of statistics refers to the field or discipline of study. In this sense of
the word, statistics is defined as follows.
Definition
Statistics Statistics is a group of methods used to collect, analyze, present, and interpret data
and to make decisions.
Every day we make decisions that may be personal, business related, or of some other kind.
Usually these decisions are made under conditions of uncertainty. Many times, the situations or
problems we face in the real world have no precise or definite solution. Statistical methods help
us make scientific and intelligent decisions in such situations. Decisions made by using statis-
tical methods are called educated guesses. Decisions made without using statistical (or scien-
tific) methods are pure guesses and, hence, may prove to be unreliable. For example, opening
a large store in an area with or without assessing the need for it may affect its success.
Like almost all fields of study, statistics has two aspects: theoretical and applied. Theoreti-
cal or mathematical statistics deals with the development, derivation, and proof of statistical
theorems, formulas, rules, and laws. Applied statistics involves the applications of those theo-
rems, formulas, rules, and laws to solve real-world problems. This text is concerned with ap-
plied statistics and not with theoretical statistics. By the time you finish studying this book, you
will have learned how to think statistically and how to make educated guesses.
1.2 Types of Statistics
Broadly speaking, applied statistics can be divided into two areas: descriptive statistics and inf-
erential statistics.
'"The Numbers Racket: How Polls and Statistics Lie," U.S. News & World Report, July 11, 1988, pp. 44-47.
USA TODAY Snapshots®
2008 U.S. patent leaders
Top five companies with most
patents earned in 2008:
2008 U.S.
PATENT
LEADERS
1 - First company en break 4,000 Issuances In one year
Source: IFI Paten: Intelligence
By Jac Vang and Sam Ward, USA TODAY
The accompanying chart shows the top five companies in the United States with the most patents. This Source. USA TODAY, January 21,
., „./,-. j , , 2009. Copyright © 2009, USA
chart describes the data on patents as collected from these five companies and, hence, is an example of today. Chart reproduced with
descriptive statistics. permission.
1.2.1 Descriptive Statistics
Suppose we have information on the test scores of students enrolled in a statistics class. In sta-
tistical terminology, the whole set of numbers that represents the scores of students is called a
data set, the name of each student is called an element, and the score of each student is called
an observation. (These terms are defined in more detail in Section 1.4.)
A data set in its original form is usually very large. Consequently, such a data set is not
very helpful in drawing conclusions or making decisions. It is easier to draw conclusions from
summary tables and diagrams than from the original version of a data set. So, we reduce data
to a manageable size by constructing tables, drawing graphs, or calculating summary measures
such as averages. The portion of statistics that helps us do this type of statistical analysis is
called descriptive statistics.
Definition
Descriptive Statistics Descriptive statistics consists of methods for organizing, displaying, and
describing data by using tables, graphs, and summary measures.
Both Chapters 2 and 3 discuss descriptive statistical methods. In Chapter 2, we learn how
to construct tables and how to graph data. In Chapter 3, we learn to calculate numerical sum-
mary measures, such as averages.
Case Study 1-1 presents an example of descriptive statistics.
1 .2.2 Inferential Statistics
In statistics, the collection of all elements of interest is called a population. The selection of a
few elements from this population is called a sample. (Population and sample are discussed in
more detail in Section 1.3.)
A major portion of statistics deals with making decisions, inferences, predictions, and fore-
casts about populations based on results obtained from samples. For example, we may make some
decisions about the political views of all college and university students based on the political
TV COM-
MERCIALS
AND
HOLIDAY
SHOPPING
USA TODAY Snapshots®
TV commercials and
holiday shopping
Did TV commercials
motivate you to shop at
a particular retailer?
Yes
^■18%
No impact,
I regularly shop there
No
44%
source: National Retail Federation survey of S.S50 adults
18 and older Margin of emir ±1 percentage point
By Jae Yang and Veronica Salozar, USA TODAY
The accompanying chart shows the degree to which TV commercials motivate people to shop at specific
retailers. According to this survey conducted by the National Retail Federation, 18% of adults included in
the survey said that they are influenced by such commercials, 30% said that they are not influenced because
they regularly shop at those stores, and 44% said that they are not influenced at all. The chart indicates
that there is ±1% margin of error. We will discuss the concept of margin of error in Chapter 8. Just to give
a quick and brief explanation, the margin of error means that the percentages given in the chart can change
in the plus or minus direction by 1% when applied to the population.
Source: USA TODAY, January 20, 2009. Copyright © 2009, USA TODAY. Chart reproduced with permission.
views of 1000 students selected from a few colleges and universities. As another example, we may
want to find the starting salary of a typical college graduate. To do so, we may select 2000 recent
college graduates, find their starting salaries, and make a decision based on this information. The
area of statistics that deals with such decision-making procedures is referred to as inferential
statistics. This branch of statistics is also called inductive reasoning or inductive statistics.
Definition
Inferential Statistics Inferential statistics consists of methods that use sample results to help
make decisions or predictions about a population.
Case Study 1-2 presents an example of inferential statistics. It shows the results of a sur-
vey in which people were asked whether or not TV commercials motivate them to shop at
specific retailers.
Chapters 8 through 13 and parts of Chapter 7 deal with inferential statistics.
Probability, which gives a measurement of the likelihood that a certain outcome will occur,
acts as a link between descriptive and inferential statistics. Probability is used to make statements
about the occurrence or nonoccurrence of an event under uncertain conditions. Probability and prob-
ability distributions are discussed in Chapters 4 through 6 and parts of Chapter 7.
EXERCISES
CONCEPTS AND PROCEDURES
1.1 Briefly describe the two meanings of the word statistics.
1.2 Briefly explain the types of statistics.
1 .3 Population Versus Sample 5
1.3 Population Versus Sample
We will encounter the terms population and sample on almost every page of this text. 2 Conse-
quently, understanding the meaning of each of these two terms and the difference between them
is crucial.
Suppose a statistician is interested in knowing
1. The percentage of all voters in a city who will vote for a particular candidate in an election
2. The 2009 gross sales of all companies in New York City
3. The prices of all houses in California
In these examples, the statistician is interested in all voters, all companies, and all houses.
Each of these groups is called the population for the respective example. In statistics, a popu-
lation does not necessarily mean a collection of people. It can, in fact, be a collection of people
or of any kind of item such as houses, books, television sets, or cars. The population of inter-
est is usually called the target population.
Definition
Population or Target Population A population consists of all elements — individuals, items, or
objects — whose characteristics are being studied. The population that is being studied is also
called the target population.
Most of the time, decisions are made based on portions of populations. For example, the
election polls conducted in the United States to estimate the percentages of voters who favor
various candidates in any presidential election are based on only a few hundred or a few thou-
sand voters selected from across the country. In this case, the population consists of all regis-
tered voters in the United States. The sample is made up of a few hundred or few thousand vot-
ers who are included in an opinion poll. Thus, the collection of a few elements selected from a
population is called a sample.
Definition
Sample A portion of the population selected for study is referred to as a sample.
Figure 1 . 1 illustrates the selection of a sample from a population.
Figure 1.1 Population and sample.
2 To learn more about sampling and sampling techniques, refer to Appendix A.
6 Chapter 1 Introduction
The collection of information from the elements of a population or a sample is called a
survey. A survey that includes every element of the target population is called a census. Often
the target population is very large. Hence, in practice, a census is rarely taken because it is
expensive and time-consuming. In many cases, it is even impossible to identify each element
of the target population. Usually, to conduct a survey, we select a sample and collect the re-
quired information from the elements included in that sample. We then make decisions based
on this sample information. Such a survey conducted on a sample is called a sample survey.
As an example, if we collect information on the 2009 incomes of all families in Connecticut,
it will be referred to as a census. On the other hand, if we collect information on the 2009
incomes of 50 families from Connecticut, it will be called a sample survey.
Definition
Census and Sample Survey A survey that includes every member of the population is called a
census. The technique of collecting information from a portion of the population is called a
sample sun'ey.
Case Study 1-3 presents an example of a sample survey.
The purpose of conducting a sample survey is to make decisions about the corresponding
population. It is important that the results obtained from a sample survey closely match the re-
sults that we would obtain by conducting a census. Otherwise, any decision based on a sample
survey will not apply to the corresponding population. As an example, to find the average income
of families living in New York City by conducting a sample survey, the sample must contain
families who belong to different income groups in almost the same proportion as they exist in
the population. Such a sample is called a representative sample. Inferences derived from a rep-
resentative sample will be more reliable.
Definition
Representative Sample A sample that represents the characteristics of the population as closely
as possible is called a representative sample.
A sample may be random or nonrandom. In a random sample, each element of the pop-
ulation has a chance of being included in the sample. However, in a nonrandom sample this
may not be the case.
Definition
Random Sample A sample drawn in such a way that each element of the population has a
chance of being selected is called a random sample. If all samples of the same size selected from
a population have the same chance of being selected, we call it simple random sampling. Such
a sample is called a simple random sample.
One way to select a random sample is by lottery or draw. For example, if we are to select
5 students from a class of 50, we write each of the 50 names on a separate piece of paper. Then
we place all 50 slips in a box and mix them thoroughly. Finally, we randomly draw 5 slips from
the box. The 5 names drawn give a random sample. On the other hand, if we arrange all 50
names alphabetically and then select the first 5 names on the list, it is a nonrandom sample be-
cause the students listed 6th to 50th have no chance of being included in the sample.
A sample may be selected with or without replacement. In sampling with replacement,
each time we select an element from the population, we put it back in the population before we
Percentage of adults who said
they had done these things in
the previous 30 days:
Survey results
ON THE
ROAD, IT'S
"DO AS I
SAY, NOT
AS I DO
Exceeded the speed limit by
1 5 mph on major highways
[5%
Exceeded the speed limit
by 15 mph on neighbor-
hood streets
15%
Deliberately ran red lights
■ 6%
Source: AAA Foundation for Traffic
Safely Oct.25-Jan. 10 survey or2.509
adu]rs.
By Alejandro Gonzalez. USA TODAY
Most motorists in the USA— 78%— call aggressive driving a serious concern yet nearly half admit speeding
on major highways in the past 30 days, according to a survey and analysis of research made public today
by the AAA Foundation for Traffic Safety.
Drivers also confessed to recently speeding on residential streets, speeding up to beat yellow lights,
honking at other drivers and tailgating, a AAA Foundation survey of 2509 adults found.
The report reflects what the Washington, D.C.-based, not-for-profit arm of the automobile club calls
American drivers' "do as I say, not as I do" attitude.
Changing that attitude is the first step toward making roads safe from aggressive driving, says Peter
Kissinger, president and CEO of the foundation, who notes that traffic crashes kill someone every 13 minutes.
Aggressive driving is a factor in 56% of fatal crashes, according to a Foundation study of federal data of
fatal crashes from 2003 through 2007, the most recent data available.
"We count (traffic deaths) in ones, twos and threes, as opposed to a plane falling out of the sky,
which gets major attention." Kissinger says, "But this is a major public health crisis."
Aggressive driving, the National Highway Traffic Safety Administration says, occurs when "an individ-
ual commits a combination of moving traffic offenses so as to endanger other persons or property." These
often include speeding, tailgating, improper lane changes, failing to yield the right of way, improper pass-
ing, and running red lights.
In the past 10 years, 14 states have taken steps against aggressive driving, according to the Gover-
nors Highway Safety Association.
"Traffic Congestion is generally the No. 1 cause for aggressive driving," says Thomas Gianni, deputy
chief of the Maryland Highway Safety Office and regional coordinator of a crackdown on aggressive driv-
ing by 150 police agencies in Maryland, Virginia, and Washington, D.C. "People are trying to get to too
many places and they don't allow themselves the time to get where they need to be."
Amanda Cooke, 21, a computer teacher in Running Springs, Calif., says she used to drive so aggres-
sively that her boyfriend was afraid to ride with her. "I'd cut people off to get into the lane I wanted to
get in," she says. "I'd tailgate them if they were going too slow or blink my lights if it was night." Cooke
says she stopped driving that way after crashing into another driver. "I didn't think it was as risky as it
was," she says.
Source: Larry Copeland, USA TODAY, April 21, 2009. Reproduced with permission.
7
8 Chapter 1 Introduction
select the next element. Thus, in sampling with replacement, the population contains the same
number of items each time a selection is made. As a result, we may select the same item more
than once in such a sample. Consider a box that contains 25 marbles of different colors.
Suppose we draw a marble, record its color, and put it back in the box before drawing the next
marble. Every time we draw a marble from this box, the box contains 25 marbles. This is an
example of sampling with replacement. The experiment of rolling a die many times is another
example of sampling with replacement because every roll has the same six possible outcomes.
Sampling without replacement occurs when the selected element is not replaced in the
population. In this case, each time we select an item, the size of the population is reduced by
one element. Thus, we cannot select the same item more than once in this type of sampling.
Most of the time, samples taken in statistics are without replacement. Consider an opinion poll
based on a certain number of voters selected from the population of all eligible voters. In this
case, the same voter is not selected more than once. Therefore, this is an example of sampling
without replacement.
EXERCISES
CONCEPTS AND PROCEDURES
1.3 Briefly explain the terms population, sample, representative sample, random sample, sampling with
replacement, and sampling without replacement.
1.4 Give one example each of sampling with and sampling without replacement.
1.5 Briefly explain the difference between a census and a sample survey. Why is conducting a sample
survey preferable to conducting a census?
■ APPLICATIONS
1.6 Explain whether each of the following constitutes a population or a sample.
a. Pounds of bass caught by all participants in a bass fishing derby
b. Credit card debts of 100 families selected from a city
c. Number of home runs hit by all Major League baseball players in the 2009 season
d. Number of parole violations by all 2147 parolees in a city
e. Amount spent on prescription drugs by 200 senior citizens in a large city
1.7 Explain whether each of the following constitutes a population or a sample.
a. Number of personal fouls committed by all NBA players during the 2008-2009 season
b. Yield of potatoes per acre for 10 pieces of land
c. Weekly salaries of all employees of a company
d. Cattle owned by 100 farmers in Iowa
e. Number of computers sold during the past week at all computer stores in Los Angeles
1.4 Basic Terms
It is very important to understand the meaning of some basic terms that will be used frequently
in this text. This section explains the meaning of an element (or member), a variable, an
observation, and a data set. An element and a data set were briefly defined in Section 1.2. This
section defines these terms formally and illustrates them with the help of an example.
Table 1.1 gives information on the 2007 charitable givings (in millions of U.S. dollars) by
six retail companies. We can call this group of companies a sample of six companies. Each
company listed in this table is called an element or a member of the sample. Table 1.1 con-
tains information on six elements. Note that elements are also called observational units.
Definition
Element or Member An element or member of a sample or population is a specific subject or object
(for example, a person, firm, item, state, or country) about which the information is collected.
1 .4 Basic Terms 9
An element 1
or a member J
Table 1.1 Charitable Givings of Six Retailers in 2007
iUU / l^IiariLaliie VrlVlIlga
Company
(millions of dollars)
Home Depot
Macy's
35.2
|Wal-Mart |
|337.9 |<— [
Best Buy
31.8
Target
168.9
Lowe's
27.5
Variable
An observation
or measurement
Source: The Chronicle of Philanthropy.
The 2007 charitable givings in our example is called a variable. The 2007 charitable givings
is a characteristic of companies that we are investigating or studying.
Definition
Variable A variable is a characteristic under study that assumes different values for different
elements. In contrast to a variable, the value of a constant is fixed.
Other examples of variables are the incomes of households, the number of houses built in
a city per month during the past year, the makes of cars owned by people, the gross profits of
companies, and the number of insurance policies sold by a salesperson per day during the past
month.
In general, a variable assumes different values for different elements, as does the 2007 char-
itable givings of the six companies in Table 1.1. For some elements in a data set, however, the
values of the variable may be the same. For example, if we collect information on incomes of
households, these households are expected to have different incomes, although some of them
may have the same income.
A variable is often denoted by x, y, or z. For instance, in Table 1.1, the 2007 charitable giv-
ings of companies may be denoted by any one of these letters. Starting with Section 1.8, we
will begin to use these letters to denote variables.
Each of the values representing the 2007 charitable givings of the six companies in Table 1.1
is called an observation or measurement.
Definition
Observation or Measurement The value of a variable for an element is called an observation or
measurement.
From Table 1.1, the 2007 charitable givings of Wal-Mart were $337.9 million. The value
$337.9 million is an observation or a measurement. Table 1.1 contains six observations, one for
each of the six retail companies.
The information given in Table 1.1 on 2007 charitable givings of companies is called the
data or a data set.
Definition
Data Set A data set is a collection of observations on one or more variables.
10 Chapter 1 Introduction
Other examples of data sets are a list of the prices of 25 recently sold homes, test scores
of 15 students, opinions of 100 voters, and ages of all employees of a company.
EXERCISES
CONCEPTS AND PROCEDURES
1.8 Explain the meaning of an element, a variable, an observation, and a data set.
■ APPLICATIONS
1.9 The following table gives the number of dog bites reported to the police last year in six cities.
City
Number of Bites
Center City
47
Elm Grove
32
Franklin
51
Bay City
44
Oakdale
12
Sand Point
3
Briefly explain the meaning of a member, a variable, a measurement, and a data set with reference to this table.
1.10 The following table gives the state taxes (in dollars) on a pack of cigarettes for nine states as of
April 1, 2009.
State Tax
State
(in dollars)
Alaska
2.00
Iowa
1.36
Massachusetts
2.51
Missouri
.17
New Hampshire
1.33
New York
2.75
Ohio
1.25
South Carolina
.07
West Virginia
.55
Briefly explain the meaning of a member, a variable, a measurement, and a data set with reference to this table.
1.11 Refer to the data set in Exercise 1.9.
a. What is the variable for this data set?
b. How many observations are in this data set?
c. How many elements does this data set contain?
1.12 Refer to the data set in Exercise 1.10.
a. What is the variable for this data set?
b. How many observations are in this data set?
c. How many elements does this data set contain?
1.5 Types of Variables
In Section 1.4, we learned that a variable is a characteristic under investigation that assumes
different values for different elements. The incomes of families, heights of persons, gross sales
of companies, prices of college textbooks, makes of cars owned by families, number of acci-
dents, and status (freshman, sophomore, junior, or senior) of students enrolled at a university
are examples of variables.
1 .5 Types of Variables 1 1
A variable may be classified as quantitative or qualitative. These two types of variables are
explained next.
1.5.1 Quantitative Variables
Some variables (such as the price of a home) can be measured numerically, whereas others (such
as hair color) cannot. The first is an example of a quantitative variable and the second that of
a qualitative variable.
Definition
Quantitative Variable A variable that can be measured numerically is called a quantitative vari-
able. The data collected on a quantitative variable are called quantitative data.
Incomes, heights, gross sales, prices of homes, number of cars owned, and number of ac-
cidents are examples of quantitative variables because each of them can be expressed numerically.
For instance, the income of a family may be $81,520.75 per year, the gross sales for a company
may be $567 million for the past year, and so forth. Such quantitative variables may be classi-
fied as either discrete variables or continuous variables.
Discrete Variables
The values that a certain quantitative variable can assume may be countable or noncountable.
For example, we can count the number of cars owned by a family, but we cannot count the
height of a family member. A variable that assumes countable values is called a discrete vari-
able. Note that there are no possible intermediate values between consecutive values of a dis-
crete variable.
Definition
Discrete Variable A variable whose values are countable is called a discrete variable. In
other words, a discrete variable can assume only certain values with no intermediate values.
For example, the number of cars sold on any day at a car dealership is a discrete variable
because the number of cars sold must be 0, 1, 2, 3, . . . and we can count it. The number of cars
sold cannot be between and 1, or between 1 and 2. Other examples of discrete variables are
the number of people visiting a bank on any day, the number of cars in a parking lot, the num-
ber of cattle owned by a farmer, and the number of students in a class.
Continuous Variables
Some variables cannot be counted, and they can assume any numerical value between two num-
bers. Such variables are called continuous variables.
Definition
Continuous Variable A variable that can assume any numerical value over a certain interval or
intervals is called a continuous variable.
The time taken to complete an examination is an example of a continuous variable because
it can assume any value, let us say, between 30 and 60 minutes. The time taken may be 42.6
minutes, 42.67 minutes, or 42.674 minutes. (Theoretically, we can measure time as precisely as
12 Chapter 1 Introduction
we want.) Similarly, the height of a person can be measured to the tenth of an inch or to the
hundredth of an inch. However, neither time nor height can be counted in a discrete fashion.
Other examples of continuous variables are weights of people, amount of soda in a 12-ounce can
(note that a can does not contain exactly 12 ounces of soda), and yield of potatoes (in pounds)
per acre. Note that any variable that involves money is considered a continuous variable.
1.5.2 Qualitative or Categorical Variables
Variables that cannot be measured numerically but can be divided into different categories are
called qualitative or categorical variables.
Definition
Qualitative or Categorical Variable A variable that cannot assume a numerical value but can be
classified into two or more nonnumeric categories is called a qualitative or categorical variable.
The data collected on such a variable are called qualitative data.
For example, the status of an undergraduate college student is a qualitative variable because
a student can fall into any one of four categories: freshman, sophomore, junior, or senior. Other
examples of qualitative variables are the gender of a person, the brand of a computer, the opin-
ions of people, and the make of a car.
Figure 1.2 illustrates the types of variables.
Variable
r
Quantitative
Discrete
(e.g., number of
houses, cars,
accidents)
1
Continuous
(e.g., length,
age, height,
weight, time)
1
Qualitative or categorical
(e.g., make of a
computer, opinions of
people, gender)
Figure 1 .2 Types of variables.
EXERCISES
CONCEPTS AND PROCEDURES
1.13 Explain the meaning of the following terms.
a. Quantitative variable
b. Qualitative variable
c. Discrete variable
d. Continuous variable
e. Quantitative data
f. Qualitative data
■ APPLICATIONS
1.14 Indicate which of the following variables are quantitative and which are qualitative.
a. Number of persons in a family
b. Colors of cars
c. Marital status of people
d. Time to commute from home to work
e. Number of errors in a person's credit report
1 .6 Cross-Section Versus Time-Series Data
1.15 Indicate which of the following variables are quantitative and which are qualitative.
a. Number of typographical errors in newspapers
b. Monthly TV cable bills
c. Spring break locations favored by college students
d. Number of cars owned by families
e. Lottery revenues of states
1.16 Classify the quantitative variables in Exercise 1.14 as discrete or continuous.
1.17 Classify the quantitative variables in Exercise 1.15 as discrete or continuous.
1.6 Cross-Section Versus Time-Series Data
Based on the time over which they are collected, data can be classified as either cross-section
or time-series data.
1.6.1 Cross-Section Data
Cross-section data contain information on different elements of a population or sample for
the same period of time. The information on incomes of 100 families for 2009 is an example
of cross-section data. All examples of data already presented in this chapter have been cross-
section data.
Definition
Cross-Section Data Data collected on different elements at the same point in time or for the
same period of time are called cross-section data.
Table 1 . 1 is reproduced here as Table 1 .2 that shows the 2007 charitable givings of six re-
tail companies. Because this table presents data on the charitable givings of six companies for
the same period (2007), it is an example of cross-section data.
Table 1.2 Charitable Givings of Six Retailers in 2007
2007 Charitable Givings
Company (millions of dollars)
Home Depot
42
Macy's
35.2
Wal-Mart
337.9
Best Buy
31.8
Target
168.9
Lowe's
27.5
Source: The Chronicle of Philanthropy.
1 .6.2 Time-Series Data
Time-series data contain information on the same element for different periods of time. Infor-
mation on U.S. exports for the years 1983 to 2009 is an example of time-series data.
Definition
Time-Series Data Data collected on the same element for the same variable at different points
in time or for different periods of time are called time-series data.
14 Chapter 1 Introduction
The data given in Table 1.3 are an example of time-series data. This table lists the total
number of indoor movie screens in the United States for the years 2003 to 2008. Note that each
screen in each theater counts as one. For example, a movieplex with 8 screens would count as
8 toward the total number of screens.
Table 1.3 Number of Movie Screens
Total Indoor
Year
Movie Screens
2003
35,361
2004
36,012
2005
37,092
2006
37,776
2007
38,159
2008
38,198
Source: National Association for Theater Owners.
1.7 Sources of Data
The availability of accurate and appropriate data is essential for deriving reliable results. 3 Data
may be obtained from internal sources, external sources, or surveys and experiments.
Many times data come from internal sources, such as a company's personnel files or ac-
counting records. For example, a company that wants to forecast the future sales of its prod-
uct may use the data of past periods from its records. For most studies, however, all the data
that are needed are not usually available from internal sources. In such cases, one may have
to depend on outside sources to obtain data. These sources are called external sources. For in-
stance, the Statistical Abstract of the United States (published annually), which contains var-
ious kinds of data on the United States, is an external source of data.
A large number of government and private publications can be used as external sources of
data. The following is a list of some government publications.
1. Statistical Abstract of the United States
2. Employment and Earnings
3. Handbook of Labor Statistics
4. Source Book of Criminal Justice Statistics
5. Economic Report of the President
6. County & City Data Book
7. State & Metropolitan Area Data Book
8. Digest of Education Statistics
9. Health United States
10. Agricultural Statistics
Most of the data contained in these books can be accessed on Internet sites such as www.
census.gov (Census Bureau), www.bls.gov (Bureau of Labor Statistics), www.ojp.usdoj.gov/bjs
(Office of Justice Program, U.S. Department of Justice, Bureau of Justice Statistics), www.os.
dhhs.gov (U.S. Department of Health and Human Services), and www.usda.gov/nass/pubs/
agstats.htm (U.S. Department of Agriculture, Agricultural Statistics).
Besides these government publications, a large number of private publications (e.g.,
Standard & Poors' Security Owner's Stock Guide and World Almanac and Book of Facts) and
3 Sources of data are discussed in more detail in Appendix A.
1.8 Summation Notation 15
periodicals (e.g., The Wall Street Journal, USA TODAY, Fortune, Forbes, and Business Week)
can be used as external data sources.
Sometimes the needed data may not be available from either internal or external sources.
In such cases, the investigator may have to conduct a survey or experiment to obtain the
required data. Appendix A discusses surveys and experiments in detail.
EXERCISES
CONCEPTS AND PROCEDURES
1.18 Explain the difference between cross-section and time-series data. Give an example of each of these
two types of data.
1.19 Briefly describe internal and external sources of data.
■ APPLICATIONS
1.20 Classify the following as cross-section or time-series data.
a. Food bill of a family for each month of 2009
b. Number of armed robberies each year in Dallas from 1998 to 2009
c. Number of supermarkets in 40 cities on December 31, 2009
d. Gross sales of 200 ice cream parlors in July 2009
1.21 Classify the following as cross-section or time-series data.
a. Average prices of houses in 100 cities
b. Salaries of 50 employees
c. Number of cars sold each year by General Motors from 1980 to 2009
d. Number of employees employed by a company each year from 1985 to 2009
1.8 Summation Notation
Sometimes mathematical notation helps express a mathematical relationship concisely. This sec-
tion describes the summation notation that is used to denote the sum of values.
Suppose a sample consists of five literary books, and the prices of these five books are $75,
$80, $35, $97, and $88, respectively. The variable price of a book can be denoted by x. The
prices of the five books can be written as follows:
Price of the first book = x l = $75
t
Subscript of x denotes the
number of the book
Similarly,
Price of the second book = x 2 = $80
Price of the third book = x 3 = $35
Price of the fourth book = x 4 = $97
Price of the fifth book = x 5 = $88
In this notation, x represents the price, and the subscript denotes a particular book.
Now, suppose we want to add the prices of all five books. We obtain
jcj + x 2 + x 3 + x 4 + x 5 = 75 + 80 + 35 + 97 + 88 = $375
The uppercase Greek letter X (pronounced sigma) is used to denote the sum of all values.
Using 2 notation, we can write the foregoing sum as follows:
2x = X; + x 2 + x 3 + x 4 + x 5 = $375
16 Chapter 1 Introduction
The notation Xx in this expression represents the sum of all the values of x and is read as "sigma
x" or "sum of all values of x."
Using summation
notation: one variable.
■ EXAMPLE 1-1
Annual salaries (in thousands of dollars) of four workers are 75, 90, 125, and 61, respectively. Find
(a) Xx (b) (Xx) 2 (c) Xx 2
Solution Let x x , x 2 , x 3 , and x 4 be the annual salaries (in thousands of dollars) of the first,
second, third, and fourth worker, respectively. Then,
x l = 75, x 2 = 90, x 3 = 125, and x 4 = 61
(a) Xx = xi + x 2 + x 3 + x 4 = 75 + 90 + 125 + 61 = 351 = $351,000
(b) Note that (X*) 2 is the square of the sum of all x values. Thus,
(Xx) 2 = (351) 2 = 123,201
(c) The expression Xx 2 is the sum of the squares of x values. To calculate Xx 2 , we first
square each of the x values and then sum these squared values. Thus,
Xx 2 = (75) 2 + (90) 2 + (125) 2 + (61) 2
= 5625 + 8100 + 15,625 + 3721 = 33,071 ■
Using summation
notation: two variables.
■ EXAMPLE 1-2
The following table lists four pairs of m and /values:
m
12
15
20
30
f
5
9
10
16
Compute the following:
(a) Xm (b) X/ 2 (c) Xmf
(d) Xm 2 /
Solution We can write
m x = 12 m 2 = 15 m 3 = 20 m 4 = 30
/ = 5 f 2 = 9 / = 10 U = 16
(a) Xm = 12 + 15 + 20 + 30 = 77
(b) X/ 2 = (5) 2 + (9) 2 + (10) 2 + (16) 2 = 25 + 81 + 100 + 256 = 462
(c) To compute Xm/ we multiply the corresponding values of m and / and then add the
products as follows:
Xm/ = mj x + m 2 f 2 + m 3 f 3 + m 4 f 4
= 12(5) + 15(9) + 20(10) + 30(16) = 875
(d) To calculate Xm 2 / we square each m value, then multiply the corresponding m and
/values, and add the products. Thus,
Xm 2 /= (m,) 2 /, + (m 2 ) 2 / 2 + (m 3 ) 2 / + (m 4 ) 2 / 4
= (12) 2 (5) + (15) 2 (9) + (20) 2 (10) + (30) 2 (16) = 21,145
The calculations done in parts (a) through (d) to find the values of Xm, X/ 2 , Xm/ and Xm 2 /
can be performed in tabular form, as shown in Table 1.4.
Table 1.4
m
/
f
mf
m 2 f
12
5
5X5
= 25
12 X 5 = 60
12 X 12 X 5 = 720
15
9
9X9
= 81
15 X 9 = 135
15
X 15 X 9 = 2025
20
10
10 X 10
= 100
20 X 10 = 200
20
X 20 X 10 = 4000
30
16
16 X 16
= 256
30 X 16 = 480
30
X 30 X 16 = 14,400
tm = 77
V= 40
2/ 2 =
462
2m/ = 875
2m 2 / = 21,145
The columns of Table 1.4 can be explained as follows.
1. The first column lists the values of m. The sum of these values gives %m
11.
2. The second column lists the values of / The sum of this column gives 2/ = 40.
3. The third column lists the squares of the /values. For example, the first value, 25, is
the square of 5. The sum of the values in this column gives X/ 2 = 462.
4. The fourth column records products of the corresponding m and /values. For exam-
ple, the first value, 60, in this column is obtained by multiplying 12 by 5. The sum of
the values in this column gives Xmf = 875.
5. Next, the m values are squared and multiplied by the corresponding/ values. The re-
sulting products, denoted by mf, are recorded in the fifth column. For example, the
first value, 720, is obtained by squaring 12 and multiplying this result by 5. The sum
of the values in this column gives %m 2 f = 21,145.
EXERCISES
CONCEPTS AND PROCEDURES
1.22 The following table lists five pairs of m and /values.
m
5
10
17
20
25
/
1
28
6
1
64
Compute the value of each of the following:
a. 2m b. 2/ 2 c. 2m/ d. 2m 2 /
1.23 The following table lists six pairs of m and /values.
m
3
6
25
12
15
18
/
16
11
16
8
4
14
Calculate the value of each of the following:
a. 2/
b. Xr,
c. 2m/ d. 2m 2 /
1.24 The following table contains information on the NCAA Men's Basketball Championship tournament
Final Four teams for the 31 -year period from 1979 to 2009. The table shows how many teams with each
seeding qualified for the Final Four during these 31 years. For example, 53 of the 124 Final Four teams
during these 31 years were seeded number one, 27 of the 124 Final Four teams were seeded number two,
and so on. Note that none of the teams seeded number 10 qualified for the Final Four in these 31 years.
Seed
1
2
3
4
5
6
7
8
9
1
Number of Teams in Men's Final Four
53
27
15
10
5
6
1
4
1
2
Let y denote the seed and x denote the number of teams having that seed. Calculate the following:
a. 2x b. 2y c. 2xy d. %y 2 e. (2y) 2
18 Chapter 1 Introduction
1.25 The following table contains the same kind of information as the table in Exercise 1.24 but for the
NCAA Women's Basketball Championship tournament Final Four teams for 28 years from 1982 to 2009.
Seed
1 2
3
4
5
6
7
8
9
Number of Teams in Women's Final Four
59 27
12
8
1
2
1
1
1
Let y denote the seed and x denote the number of teams having that seed. Calculate the following:
a. %x b. Xy c. 2xy d. Sy 2 e. (Zyj 2
■ APPLICATIONS
1.26 Eight randomly selected customers at a local grocery store spent the following amounts on groceries
in a single visit: $216, $184, $35, $92, $144, $175, $11, and $57, respectively. Let y denote the amount
spent on groceries in a single visit. Find:
a. ty b. (2y) 2 e. Xy 2
1.27 The number of pizzas delivered to a college campus on six randomly selected nights is 48, 103, 95,
188, 286, and 136, respectively. Let x denote the number of pizzas delivered to this college campus on any
given night. Find:
a. %x b. (2jc) 2 c. %x 2
1.28 Prices (in thousands of dollars) of five new cars are 28, 35, 39, 54, and 18, respectively. Let x be the
price of a new car in this sample. Find:
a. tx b. (2jc) 2 c. tx 2
1.29 The number of students (rounded to the nearest thousand) currently enrolled at seven universities is
7, 39, 21, 16, 3, 43, and 19, respectively. Let x be the number of students currently enrolled at a univer-
sity. Find:
a. Xx b. (Zx) 2 c. Sjc 2
USES AND MISUSES... speaking the language of statistics
Have you ever heard the statistic "the average American family has
2.1 children?" What is wrong with this statement, and how do we fix
it? How about: "In a representative sample of 10 American families,
one can expect there to be 21 children." The statement is wordy but
more accurate. Why do we care?
Statisticians pay close attention to definitions because, without
them, calculations would be impossible to make and interpretations
of the data would be meaningless. Often, when you read statistics
reported in the newspaper, the journalist or editor sometimes
chooses to describe the results in a way that is easier to understand
but that distorts the actual statistical result.
Let us pick apart our example. The word average has a very spe-
cific meaning in probability (Chapters 4 and 5). The intended meaning
of the word here really is typical. The adjective American helps us de-
fine the population. The Census Bureau defines family as "a group of
two people or more (one of whom is the householder) related by birth,
marriage, or adoption and residing together; all such people (including
related subfamily members) are considered as members of one family."
It defines children as "all persons under 18 years, excluding people who
maintain households, families, or subfamilies as a reference person or
spouse." We understand implicitly that a family cannot have a fractional
number of children, so we accept that this discrete variable takes on
the properties of a continuous variable when we are talking about the
characteristics of a large population. How large does the population
need to be before we can derive continuous variables from discrete
variables? The answer comes in the chapters that follow.
The moral of the story is that whenever you read a statistical re-
sult, be sure that you understand the definitions of the terms used
to describe the result and relate those terms to the definitions that
you already know. In some cases year is a categorical variable, in oth-
ers it is a discrete variable, and in others it is a continuous variable.
Many surveys will report that "respondents feel better, the same, or
worse" about a particular subject. Although better, same, and worse
have a natural order to them, they do not have numerical values.
Glossary
Census A survey that includes all members of the population.
Continuous variable A (quantitative) variable that can assume any
numerical value over a certain interval or intervals.
Cross-section data Data collected on different elements at the
same point in time or for the same period of time.
Supplementary Exercises 19
Data or data set Collection of observations or measurements on
a variable.
Descriptive statistics Collection of methods for organizing, display-
ing, and describing data using tables, graphs, and summary measures.
Discrete variable A (quantitative) variable whose values are
countable.
Element or member A specific subject or object included in a
sample or population.
Inferential statistics Collection of methods that help make deci-
sions about a population based on sample results.
Observation or measurement The value of a variable for an element.
Population or target population The collection of all elements
whose characteristics are being studied.
Qualitative or categorical data Data generated by a qualitative
variable.
Qualitative or categorical variable A variable that cannot assume
numerical values but is classified into two or more categories.
Quantitative data Data generated by a quantitative variable.
Quantitative variable A variable that can be measured numerically.
Supplementary Exercises
1.30 The following table gives the total number of DVDs sold at retail stores between 2003 and 2008.
Year
U.S. Retail Sales of DVDs
(millions of DVDs)
2003
11.3
2004
15.1
2005
16.0
2006
16.3
2007
15.8
2008
15.2
Source: SNL Kagan.
Describe the meanings of a variable, a measurement, and a data set with reference to this table.
1.31 The following table gives the total 2009 payrolls (on the opening day of the 2009 season, rounded to
the nearest million dollars) for eight National League baseball teams.
Total Payroll
Team
(millions of dollars)
Atlanta Braves
97
Chicago Cubs
135
Florida Marlins
37
Los Angeles Dodgers
100
New York Mets
149
Philadelphia Phillies
113
Pittsburg Pirates
49
San Francisco Giants
83
Describe the meanings of a member, a variable, a measurement, and a data set with reference to this table.
1.32 Refer to Exercises 1.30 and 1.31. Classify these data sets as either cross-section or time-series data.
Random sample A sample drawn in such a way that each element
of the population has some chance of being included in the sample.
Representative sample A sample that contains the same charac-
teristics as the corresponding population.
Sample A portion of the population of interest.
Sample survey A survey that includes elements of a sample.
Simple random sampling If all samples of the same size selected
from a population have the same chance of being selected, it is called
simple random sampling. Such a sample is called a simple random
sample.
Statistics Group of methods used to collect, analyze, present, and
interpret data and to make decisions.
Survey Collection of data on the elements of a population or
sample.
Time-series data Data that give the values of the same variable
for the same element at different points in time or for different pe-
riods of time.
Variable A characteristic under study or investigation that assumes
different values for different elements.
20 Chapter 1 Introduction
1.33 Indicate whether each of the following examples refers to a population or to a sample.
a. A group of 25 patients selected to test a new drug
b. Total items produced on a machine for each year from 1995 to 2009
c. Yearly expenditures on clothes for 50 persons
d. Number of houses sold by each of the 10 employees of a real estate agency during 2009
1.34 Indicate whether each of the following examples refers to a population or to a sample.
a. Salaries of CEOs of all companies in New York City
b. Five hundred houses selected from a city
c. Gross sales for 2009 of four fast-food chains
d. Annual incomes of all 33 employees of a restaurant
1.35 State which of the following is an example of sampling with replacement and which is an example
of sampling without replacement.
a. Selecting 10 patients out of 100 to test a new drug
b. Selecting one professor to be a member of the university senate and then selecting one professor
from the same group to be a member of the curriculum committee
1.36 State which of the following is an example of sampling with replacement and which is an example
of sampling without replacement.
a. Selecting seven cities to market a new deodorant
b. Selecting a high school teacher to drive students to a lecture in March, then selecting a teacher
from the same group to chaperone a dance in April
1.37 The number of shoe pairs owned by six women is 8, 14, 3, 7, 10, and 5, respectively. Let x denote
the number of shoe pairs owned by a woman. Find:
a. tx b. (tx) 2 c. tx 2
1.38 The number of restaurants in each of five small towns is 4, 12, 8, 10, and 5, respectively. Let y de-
note the number of restaurants in a small town. Find:
a. 2y b. (2y) 2 c. 2y 2
1.39 The following table lists five pairs of m and /values.
m
3
16
11
9
20
f
7
32
17
12
34
Compute the value of each of the following:
a. tm b. 2/ 2 c. 2m/ d. 2m 2 / e. 2m 2
1.40 The following table lists six pairs of x and y values.
X
7
11
8
4
14
28
y
5
15
7
10
9
19
Compute the value of each of the following:
a. 2y b. 2x 2 c. txy d. tx 2 y e. 2y 2
Self-Review Test
1. A population in statistics means a collection of all
a. men and women
b. subjects or objects of interest
c. people living in a country
2. A sample in statistics means a portion of the
a. people selected from the population of a country
b. people selected from the population of an area
c. population of interest
3. Indicate which of the following is an example of a sample with replacement and which is a sample
without replacement.
a. Five friends go to a livery stable and select five horses to ride (each friend must choose a differ-
ent horse).
b. A box contains five marbles of different colors. A marble is drawn from this box, its color is recorded,
and it is put back into the box before the next marble is drawn. This experiment is repeated 12 times.
Mini-Project 21
4. Indicate which of the following variables are quantitative and which are qualitative. Classify the quan-
titative variables as discrete or continuous.
a. Women's favorite TV programs
b. Salaries of football players
c. Number of pets owned by families
d. Favorite breed of dog for each of 20 persons
5. The following table contains data on the 10 biggest Nasdaq losers of October 2008. The first column
in the table contains the names of the companies and their NASDAQ symbols, and the second column
gives the returns for the stocks of these companies for the month of October 2008.
Company (NASDAQ Symbol) October 2008 Return
Smurfit-Stone Container (SSCC)
-71
3%
Bruker Corporation (BRKR)
-69
3%
Savient Pharmaceuticals (SVNT)
-68
\%
Parexel International Corp (PRXL)
-63
7%
Global Industries (GLBL)
-63
3%
Rigel Pharmaceuticals (RIGL)
-62
7%
Liberty Media Interactive (LINTA)
-62
2%
YRC Worldwide (YRCW)
-61
7%
Grupo Financiero Galicia S.A. (GOAL)
-61
7%
Bare Escentuals (BARE)
-61
5%
Source: The Motley Fool.
Explain the meaning of a member, a variable, a measurement, and a data set with reference to this table.
6. The number of credit cards possessed by five couples is 2, 5, 3, 12, and 7, respectively. Let x be the
number of credit cards possessed by a couple. Find:
a. %x b. (2x) 2 c. 2x 2
7. The following table lists five pairs of m and /values.
m
3
6
9
12
15
f
15
25
40
20
12
Calculate
a. 2m b. 2/ c. 2m 2 d. 2m/ e. 2m 2 / f. 2/ 2
Mini-Project
■ MINI-PROJECT 1-1
In this mini-project, you are going to obtain a data set of interest to you that you will use for mini-projects
in some of the other chapters. The data set should contain at least one qualitative variable and one quan-
titative variable, although having two of each will be necessary in some cases. Ask your instructor how
many variables you should have. A good-size data set to work with should contain somewhere between
50 and 100 observations.
Here are some examples of the procedures to use to obtain data:
1. Take a random sample of used cars and collect data on them. You may use Web sites like Cars.com,
AutoTrader.com, and so forth. Quantitative variables may include the price, mileage, and age of a car.
Categorical variables may include the model, drive train (front wheel, rear wheel, and so forth), and
type (compact, SUV, minivan, and so forth). You can concentrate on your favorite type of car, or look
at a variety of types.
2. Examine the real estate ads in your local newspaper or online and obtain information on houses
for sale that may include listed price, number of bedrooms, lot size, living space, town, type of house,
number of garage spaces, and number of bathrooms.
3. Use an almanac or go to a government Web site, such as www.census.gov or www.cdc.gov, to ob-
tain information for each state. Quantitative variables may include income, birth and death rates, cancer
22 Chapter 1 Introduction
incidence, and the proportion of people living below the poverty level. Categorical variables may in-
clude things like the region of the country where each state is located and which party won the state
governorship in the last election. You can also collect this information on a worldwide level and use the
continent or world region as a categorical variable.
4. Take a random sample of students and ask them questions such as:
• How much money did you spend on books last semester?
• How many credit hours did you take?
• What is your major?
5. If you are a sports fan, you can use an almanac or sports Web site to obtain statistics on a random
sample of athletes. You can look at sport-specific statistics such as home runs, runs batted in, position,
left-handed/right-handed, and so forth in baseball, or you could collect information to compare differ-
ent sports by gathering information on salary, career length, weight, and so forth.
Once you have collected the information, write a brief report that includes answers to the following
tasks/questions:
a. Describe the variables on which you have collected information.
b. Describe a reasonable target population for the sample you used.
c. Is your sample a random sample from this target population?
d. Do you feel that your sample is representative of this population?
e. Is this an example of sampling with or without replacement?
f. For each quantitative variable, state whether it is continuous or discrete.
g. Describe the meaning of an element, a variable, and a measurement for this data set.
h. Describe any problems you faced in collecting these data.
i. Were any of the data values unusable? If yes, explain why.
Your instructor will probably want to see a copy of the data you collected. If you are using statisti-
cal software in the class, enter the data into that software and submit a copy of the data file. If you are
using a handheld technology calculator, such as a graphing calculator, you will probably have to print out
a hard copy version of the data set. Save this data set for projects in future chapters.
ECHNOLOGY
INSTRUCTION
Entering and Saving Data
LI
L2
L3 1
L1(15 =
Screen 1.1
LI
LI
75
EH
E3
S3
E2
12
£1
31
50
20
19
L2(1)=50
Screen 1.2
Technology makes the process of data analysis much easier and faster. Therefore, you need to be able to enter
the data, proofread them, and revise them. Moreover, you can save the data and retrieve them for use at a
later date.
Entering Data in a List
1. On the TI-84, variables are referred to as lists.
2. In order to enter data into the TI-84, you first need to decide whether you want to save the
data for later use or just use it in the immediate future.
3. If you will be just using it in the immediate future, select STAT >EDIT >SetUpEditor,
and then press Enter. This will set up the editor to use "scratch" lists LI, L2, L3, L4,
L5, and L6 (Screen 1.1). Now select STAT >EDIT >Edit and start typing your numeric
data into the column or columns, pressing ENTER after each entry (Screen 1.2). Note
that the TI-84 calculator will not handle nonnumeric data.
Changing List Names/Establishing Visible Lists
1. The TI-84 has only six "scratch" lists. In some cases you will be using your data at a later
date. You can rename a list so that you do not have to reenter the data. Select STAT
>EDIT >SetUpEditor, and then type in the names of your variables separated by commas
(Screens 1.3 and 1.4). Names can be one to five letters long, with the letters found in
Technology Instructi on 2 5
green on your keypad. You can use the green ALPHA key with each letter, or press
A-LOCK (2nd > ALPHA) while you are typing the name. To turn off A-LOCK, press
ALPHA.
2. You can use the arrow keys to move around and go back to a cell to edit its contents.
When editing values, you will need to press ENTER for the changes to take effect.
Screen i .3 3. SetUpEditor determines what lists are displayed in the editor. Changing what SetUpEditor
displays does not delete any lists. Your lists remain in storage when the calculator is
turned off.
Numeric Operations on Lists
1. To calculate the sum of the values in a list, such as LI, select LIST (2nd > STAT) >
MATH > sum(. Enter the name of the list (e.g., 2nd > 1 for LI), then type the right
parenthesis. Press ENTER. (See Screens 1.5 and 1.6.)
2. If you need to find the sum of values and the square of the sum denoted by (Sx) 2 , you
can use the same instructions as in item 1. However, just before you press ENTER,
press the x 2 button. If you wish to square each value and calculate the sum of the
squared values, which is denoted by Sjt, press the x 2 button after entering the name of
the list but prior to typing the right parenthesis. Screen 1.6 shows the appearance of
these two processes.
SetUpEditor EX1,
EX2
□one
EHi
EK2
i
?E
GH
S3
12
El
31
50
20
Hi
EK2C1) =50
NRMES OPS mac
1 : min<
2:max<
3: meant;
4: nedianC
iflsupi<
bTprodC
74stdDeu<
sum(l_i >
2S5
sum(l_i > £
81225
sun i. L 1 E >
15655
I
Screen 1.5
Screen 1.6
Entering and Saving Data
Start Minitab. You will see the computer screen divided into two parts — Session
window, which will contain numeric output; a Worksheet, which looks similar to a
spreadsheet, where you will enter your data (see Screen 1.7); and a Project Manager
window. You are allowed to have multiple worksheets within a project.
+
C1
C?
C3-T
year
sales
employee
1
2008
35
J. Smith
2
2008
38
A. Jones
3
2009
50
J. Smith
4
2009
48
A. Jones
Screen 1.7
2. Use the mouse or the arrow keys to select where you want to start entering your data.
Each column in the worksheet corresponds to a variable, so you can enter only one kind
of data into a given column. Data can be numeric, text, or date/time. The rectangles in
the worksheet are called cells, and the cells are organized into columns such as CI, C2,
and so on, each with rows 1, 2, and so forth. Note that if a column contains text data,
"-T" will be added to the column heading.
3. The blank row between the column labels and the first row is for variable names. In
these blank cells, you can type the names of variables.
4. You can change whether you are typing the data across in rows or down in columns
by clicking the direction arrow at the top left of the worksheet (also shown in
Screen 1.7).
24 Chapter 1 Introduction
5. Click on a cell and begin typing. Press Enter when you are finished with that cell.
6. If you need to revise an entry, go to that cell with the mouse or the arrow keys and begin
typing. Press Enter to put the revised entry into the cell.
7. When you are done, select File >Save Project As to save your work for the first time as
a file on your computer. Note that Minitab will automatically assign the file extension
.mpj to your work after you choose the filename.
8. Try entering the following data into Minitab:
January
52
.08
February
48
.06
March
49
.07
Name the columns Month, Sales, Increase. Save the result as the file test.mpj.
9. To retrieve the file, select File >Open and select the file test.mpj.
10. If you are already in Minitab and you want to start a new worksheet, select File >New
and choose Worksheet. Whenever you save a project, Minitab will automatically save all
of the worksheets in the project.
Calculator
ci
C3 employee
Select
Help
Screen 1.8
Creating New Columns from Existing Columns
In some circumstances, such as when you need to calculate Sx 2 or 2xy, you will need to
calculate a new column of values using one or more existing columns. To calculate a column
containing the squares of the values in the column Sales as shown in Screen 1.7,
Store result in variable: | c4
Expression:
'sales'**2
Functions:
.7 i Bj 9j t j _-:<>:
All functions
d
Absolute value
Antilog
Any
; : l
And ;
Or |
Arcsine
Arccosine
Arctangent
V
Not |
Select
l~~ Assign as a formula
OK
Cancel
1. Select Calc > Calculator.
2. Type the name of the column to contain the new values
(such as C4) in the Store result in variable: box.
3. Click inside the Expression: box, click C2 Sales in the
column to the left of the Expression: box, and click Select.
Click on the exponentiation (**) button. Type 2 after the
two asterisks in the Expression: box. Click OK. (See
Screen 1.8.)
4. The numbers 1225, 1444, 2500, and 2304 should appear
in C4.
Calculating the Sum of a Column
1. To calculate the sum of the values in a column, select Calc > Column Statistics, which
will produce a dialog box. From the Statistic list, select Sum.
2. Click in the Input Variable: box. The list of variables will appear in the left portion of the
dialog box. Click on the variable you wish to sum, then click Select. (See Screen 1.9.)
Technology Instruction 25
Column Statistics
Cl year
C2 sales
C4
Statistic
[* Sum
C Mean
C Standard deviation
C Minimum
f* Maximum
C Range
C Median
(~ Sum of squares
r N total
f~ N nonmissing
i™ N missing
Input variable: [c4~
Store result in: f
Select
Help
Screen 1.9
OK
(Optional)
Cancel
Session f^~)(~n)
Sum of C4
Sun of C4 = 7473
Screen 1.10
3. Click OK. The result will appear in the Session window. (See Screen 1.10.)
Entering and Saving Data in Excel
1. Start Excel.
2. Use the mouse or the arrow keys to select where you want to start entering your data.
Data can be numeric or text. The rectangles are called cells, and the cells are collectively
known as a spreadsheet.
3. You can format your data by selecting the cells that you want to format, then selecting
Format > Cells, and then choosing whether you want to format a number, align text, and
so forth. For common formatting tasks, you have icons on the toolbar, such as a dollar
sign ($) to format currency, a percent sign (%) to format numbers as percents, and icons
representing left-, center-, and right-aligned text to change your alignment.
4. If you need to revise an entry, go to that cell with the mouse or the arrow keys. You can
retype the entry or you can edit it. To edit it, double-click on the cell and use the arrow
keys and the backspace key to help you revise the entry, then press Enter to put the revised
entry into the cell.
5. When you are done, select File >Save As to save your work for the first time as a file on
your computer. Note that Excel will automatically assign the file extension .xls to your
work after you choose the filename.
6. Try entering the following data into Excel:
Format it to look like this:
January
52
.08
February
48
.06
March
49
.07
January
$52.00
8%
February
$48.00
6%
March
$49.00
7%
Save the result as the file test.xls.
26 Chapter 1 Introduction
■
Mi
A
B
c
1
year
sales
employee
2
2008
35 J. Smith
3
2008
38 A. Jones
4
2009
50 J. Smith
5
2009
48 A. Jones
7. To retrieve the file, select File > Open and select the file test.xls.
Screen 1.11 contains the data from the Minitab example (Screen 1.7) as displayed in
Excel.
Screen 1.11
SUM
T (j» X ✓ fr =B1 A 2
A
B
c
D
1
January
$52.00
■ — — 1
s%
=Bl*2
2
February
$48.00
6%
3
March
$49.00
7%
4
Screen 1.12
Creating New Columns from Existing Columns
Many times, such as when you need to calculate Sjt 2 or Xxy, you will need to calculate a new
column of values using one or more existing columns. To calculate the squares of the values in
cells Bl to B3 and place them in cells Dl to D3:
1. Click on cell Dl.
2. Type =B1 A 2. Press Enter. (See Screen 1.12.)
While still on cell Dl, select Edit > Copy. Highlight cells D2 and D3.
Select Edit > Paste.
3. The numbers 2704, 2304, and 2401 should appear in Dl to D3.
Calculating the Sum of a Column
SUM
(j> X Ji =SUM(D1:D3}
A
B
C
E
F
1
January
$52.00
s%
2704
2
February
$4S.O0
6%
2304
3
March
$49.00
7%
2401
4
=SUM(B, ^)
5
| SUMinumberl, [number^]. ..,] |
To calculate the sum of the values in a column, go to the
empty cell below the values you wish to find the sum of.
Click the sigma (X) button in the upper-right portion of the
Home tab. This will enter the Sum function into the cell
along with the list of cells involved in the sum. (Note: If the
list is incorrect, you can type any changes.) Press Enter. (See
Screen 1.13.)
Screen 1.13
TECHNOLOGY ASSIGNMENTS
TAl.l The following table gives the names, hours worked, and salary for the past week for five workers.
Name
Hours Worked
Salary ($)
John
42
925
Shannon
33
2583
Kathy
28
1255
David
47
2090
Steve
40
1020
a. Enter these data into the spreadsheet. Save the data file as WORKER. Exit the session or program.
Then restart the program or software and retrieve the file WORKER.
b. Print a hard copy of the spreadsheet containing data you entered.
TA1.2 Refer to data on 2007 charitable givings of six retailers given in Table 1.1. Enter those data into
the spreadsheet and save this file as GIVINGS.
Chapter
Organizing and Graphing Data
What future careers interest high school students the most? Information technology? Business/
management? Health care? Or some other careers? A sample survey of high school students showed
the percentage of students intending to enter each of these fields. In this sample of 1023 high school
students, 15% said that they planned to pursue a career in health care. (See Case Study 2-1)
In addition to thousands of private organizations and individuals, a large number of U.S. government
agencies (such as the Bureau of the Census, the Bureau of Labor Statistics, the National Agricultural
Statistics Service, the National Center for Education Statistics, the National Center for Health Statistics,
and the Bureau of Justice Statistics) conduct hundreds of surveys every year. The data collected from
each of these surveys fill hundreds of thousands of pages. In their original form, these data sets may
be so large that they do not make sense to most of us. Descriptive statistics, however, supplies the
techniques that help to condense large data sets by using tables, graphs, and summary measures. We
see such tables, graphs, and summary measures in newspapers and magazines every day. At a glance,
these tabular and graphical displays present information on every aspect of life. Consequently, descriptive
statistics is of immense importance because it provides efficient and effective methods for summa-
rizing and analyzing information.
This chapter explains how to organize and display data using tables and graphs. We will learn
how to prepare frequency distribution tables for qualitative and quantitative data; how to construct bar
graphs, pie charts, histograms, and polygons for such data; and how to prepare stem-and-leaf displays.
2.1 Raw Data
2.2 Organizing and Graphing
Qualitative Data
Case Study 2-1 Career
Choices for High School
Students
Case Study 2-2 In or Out in
30 Minutes
2.3 Organizing and Graphing
Quantitative Data
Case Study 2-3 Morning
Grooming
2.4 Shapes of Histograms
2.5 Cumulative Frequency
Distributions
2.6 Stem-and-Leaf Displays
2.7 Dotplots
27
28 Chapter 2 Organizing and Graphing Data
2.1 Raw Data
When data are collected, the information obtained from each member of a population or sample
is recorded in the sequence in which it becomes available. This sequence of data recording is
random and unranked. Such data, before they are grouped or ranked, are called raw data.
Definition
Raw Data Data recorded in the sequence in which they are collected and before they are
processed or ranked are called raw data.
Suppose we collect information on the ages (in years) of 50 students selected from a uni-
versity. The data values, in the order they are collected, are recorded in Table 2. 1 . For instance,
the first student's age is 21, the second student's age is 19 (second number in the first row), and
so forth. The data in Table 2.1 are quantitative raw data.
Table 2.1 Ages of 50 Students
21
19
24
25
29
34
26
27
37
33
18
20
19
22
19
19
25
22
25
23
25
19
31
19
23
18
23
19
23
26
22
28
21
20
22
22
21
20
19
21
25
23
18
37
27
23
21
25
21
24
Suppose we ask the same 50 students about their student status. The responses of the students
are recorded in Table 2.2. In this table, F, SO, J, and SE are the abbreviations for freshman,
sophomore, junior, and senior, respectively. This is an example of qualitative (or categorical)
raw data.
Table 2.2 Status of 50 Students
J
F
SO
SE
J
J
SE
J
J
J
F
F
J
F
F
F
SE
SO
SE
J
J
F
SE
SO
SO
F
J
F
SE
SE
SO
SE
J
SO
SO
J
J
SO
F
SO
SE
SE
F
SE
J
SO
F
J
SO
SO
The data presented in Tables 2.1 and 2.2 are also called ungrouped data. An ungrouped
data set contains information on each member of a sample or population individually.
2.2 Organizing and Graphing Qualitative Data
This section discusses how to organize and display qualitative (or categorical) data. Data sets
are organized into tables, and data are displayed using graphs.
2.2.1 Frequency Distributions
A sample of 100 students enrolled at a university were asked what they intended to do after
graduation. Forty-four said they wanted to work for private companies/businesses, 16 said
they wanted to work for the federal government, 23 wanted to work for state or local governments,
2.2 Organizing and Graphing Qualitative Data 29
and 17 intended to start their own businesses. Table 2.3 lists the types of employment and the
number of students who intend to engage in each type of employment. In this table, the vari-
able is the type of employment, which is a qualitative variable. The categories (representing
the type of employment) listed in the first column are mutually exclusive. In other words,
each of the 100 students belongs to one and only one of these categories. The number of stu-
dents who belong to a certain category is called the frequency of that category. A frequency
distribution exhibits how the frequencies are distributed over various categories. Table 2.3
is called a frequency distribution table or simply a frequency table.
Variable ■
Category -
Table 2.3 Type of Employment Students
Intend to Engage In
Type of Employment
Private companies/businesses
Federal government
State/local government
Own business
Number of
Students
44
m
23
17
Sum = 100
Frequency column
Frequency
Definition
Frequency Distribution for Qualitative Data A frequency distribution for qualitative data lists all
categories and the number of elements that belong to each of the categories.
Example 2-1 illustrates how a frequency distribution table is constructed for qualitative data.
EXAMPLE 2-1
A sample of 30 employees from large companies was selected, and these employees were
asked how stressful their jobs were. The responses of these employees are recorded below,
where very represents very stressful, somewhat means somewhat stressful, and none stands
for not stressful at all.
somewhat
very
very
somewhat
somewhat
none
somewhat
somewhat
very
very
somewhat
somewhat
none
somewhat
very
very
very
very
somewhat
somewhat
very
somewhat
none
very
none
none
somewhat
somewhat
none
somewhat
Construct a frequency distribution table for these data.
Solution Note that the variable in this example is how stressful is an employee's job. This
variable is classified into three categories: very stressful, somewhat stressful, and not stressful at
all. We record these categories in the first column of Table 2.4. Then we read each employee's
response from the given data and mark a tally, denoted by the symbol |, in the second column
of Table 2.4 next to the corresponding category. For example, the first employee's response is
that his or her job is somewhat stressful. We show this in the frequency table by marking a tally
in the second column next to the category somewhat. Note that the tallies are marked in blocks
of five for counting convenience. Finally, we record the total of the tallies for each category in
the third column of the table. This column is called the column of frequencies and is usually de-
noted by / The sum of the entries in the frequency column gives the sample size or total fre-
quency. In Table 2.4, this total is 30, which is the sample size.
Constructing a
frequency distribution table
for qualitative data.
30 Chapter 2 Organizing and Graphing Data
Table 2.4 Frequency Distribution of Stress on Job
Stress on Job
Tally
Frequency (/)
Very
mm
10
Somewhat
mini
14
None
mi
6
Sum = 30
2.2.2 Relative Frequency and Percentage Distributions
The relative frequency of a category is obtained by dividing the frequency of that category by
the sum of all frequencies. Thus, the relative frequency shows what fractional part or propor-
tion of the total frequency belongs to the corresponding category. A relative frequency distri-
bution lists the relative frequencies for all categories.
Calculating Relative Frequency of a Category
Relative frequency of a category =
Frequency of that category
Sum of all frequencies
The percentage for a category is obtained by multiplying the relative frequency of that cat-
egory by 100. A percentage distribution lists the percentages for all categories.
Calculating Percentage
Percentage = (Relative frequency) • 100
■ EXAMPLE 2-2
Determine the relative frequency and percentage distributions for the data of Table 2.4.
Solution The relative frequencies and percentages from Table 2.4 are calculated and listed
in Table 2.5. Based on this table, we can state that .333, or 33.3%, of the employees said that
their jobs are very stressful. By adding the percentages for the first two categories, we can
state that 80% of the employees said that their jobs are very or somewhat stressful. The other
numbers in Table 2.5 can be interpreted the same way.
Notice that the sum of the relative frequencies is always 1.00 (or approximately 1.00 if the
relative frequencies are rounded), and the sum of the percentages is always 100 (or
approximately 100 if the percentages are rounded).
Table 2.5 Relative Frequency and Percentage Distributions
of Stress on Job
Stress on Job
Relative Frequency
Percentage
Very
10/30 = .333
.333(100) = 33.3
Somewhat
14/30 = .467
.467(100) = 46.7
None
6/30 = .200
.200(100) = 20.0
firm
Constructing relative
frequency and percentage
distributions.
Sum = 1.000
Sum = 100
USA TODAY Snapshots®
Career choices for
high school students
Source: RJdgfd survey of 1,033 high
y jdCS ^hool students 14 to 18 years old.
Margin of error ±1 percentage points
CAREER
CHOICES
FOR HIGH
SCHOOL
STUDENTS
By Ja e Yang .and Sam Wa rd , USA TODAY
The accompanying chart shows a bar graph indicating the career choices of high school students. Note that
in this chart, the bars are drawn horizontally. The chart shows the percentage of high school students in a
sample of 1023 who plan to go into different careers. Note that the percentages in the chart add up to 72%.
Therefore, 28% of these students mentioned other careers.
Source: USA Today, April 1 6, 2009.
Copyright © 2009, USA Today. Chart
reproduced with permission.
2.2.3 Graphical Presentation of Qualitative Data
All of us have heard the adage "a picture is worth a thousand words." A graphic display can re-
veal at a glance the main characteristics of a data set. The bar graph and the pie chart are two
types of graphs that are commonly used to display qualitative data.
Bar Graphs
To construct a bar graph (also called a bar chart), we mark the various categories on the hor-
izontal axis as in Figure 2.1. Note that all categories are represented by intervals of the same
width. We mark the frequencies on the vertical axis. Then we draw one bar for each category
such that the height of the bar represents the frequency of the corresponding category. We leave
a small gap between adjacent bars. Figure 2.1 gives the bar graph for the frequency distribution
of Table 2.4.
15
» 12
o
§ 9
cr
<u
£ 6
n
Definition
Bar Graph A graph made of bars whose heights represent the frequencies of respective categories
is called a bar graph.
The bar graphs for relative frequency and percentage distributions can be drawn simply
by marking the relative frequencies or percentages, instead of the frequencies, on the verti-
cal axis.
Sometimes a bar graph is constructed by marking the categories on the vertical axis and
the frequencies on the horizontal axis. Case Study 2-1 presents such an example.
Stress on job
Figure 2.1 Bar graph
for the frequency distri-
bution of Table 2.4.
31
IN OR
OUT IN 30
MINUTES
USA TODAY Snapshots®
In or out in 30 minutes
How long do you take
to make your decision
on candidates after
interviews?
Less than
30 minutes
47%
More than „
an hour 30 minutes
to an hour
19%
34%
Source; Development Dimensions
International (DDI) survey or 1 .91
interviewers. Margin of error ±3
percentage points.
By Jae Yang jnd Alejandro Gonzalez. USA TODAY
Source. USA Today, March 19, T h e accompanying chart shows a pie chart indicating the time it takes interviewers to make decisions after
2009. Copyright © 2009, USA , . . ,. , , , , . . . , ,
Today. Chart reproduced with tnev lnterview candidates. For example, 47% of the interviewers said that they decide within 30 minutes
permission. after interviewing a candidate. The pie chart is based on a sample survey of 1910 interviewers.
Pie Charts
A pie chart is more commonly used to display percentages, although it can be used to display
frequencies or relative frequencies. The whole pie (or circle) represents the total sample or popu-
lation. Then we divide the pie into different portions that represent the different categories.
Definition
Pie Chart A circle divided into portions that represent the relative frequencies or percentages
of a population or a sample belonging to different categories is called a pie chart.
As we know, a circle contains 360 degrees. To construct a pie chart, we multiply 360 by
the relative frequency of each category to obtain the degree measure or size of the angle for the
corresponding category. Table 2.6 shows the calculation of angle sizes for the various categories
of Table 2.5.
Table 2.6 Calculating Angle Sizes for the Pie Chart
Stress on Job
Relative Frequency
Angle Size
Very
.333
360(.333) = 119.88
Somewhat
.467
360(.467) = 168.12
None
.200
360(.200) = 72.00
Sum = 1.000
Sum = 360
Figure 2.2 shows the pie chart for the percentage distribution of Table 2.5, which uses the
angle sizes calculated in Table 2.6.
32
2.2 Organizing and Graphing Qualitative Data
33
Figure 2.2 Pie chart for the percentage distribution
of Table 2.5.
EXERCISES
CONCEPTS AND PROCEDURES
2.1 Why do we need to group data in the form of a frequency table? Explain briefly.
2.2 How are the relative frequencies and percentages of categories obtained from the frequencies of
categories? Illustrate with the help of an example.
2.3 The following data give the results of a sample survey. The letters A, B, and C represent the three
categories.
A
B
B
A
C
B
C
C
C
A
C
B
C
A
c
C
B
c
c
A
A
B
C
C
B
C
B
A
c
A
a. Prepare a frequency distribution table.
b. Calculate the relative frequencies and percentages for all categories.
c. What percentage of the elements in this sample belong to category B?
d. What percentage of the elements in this sample belong to category A or C?
e. Draw a bar graph for the frequency distribution.
2.4 The following data give the results of a sample survey. The letters Y, N, and D represent the three
categories.
D
N
N
Y
Y
Y
N
Y
D
Y
Y
Y
Y
Y
N
Y
Y
N
N
Y
N
Y
Y
N
D
N
Y
Y
Y
Y
Y
Y
N
N
Y
Y
N
N
D
Y
a. Prepare a frequency distribution table.
b. Calculate the relative frequencies and percentages for all categories.
c. What percentage of the elements in this sample belong to category Y?
d. What percentage of the elements in this sample belong to category N or D?
e. Draw a pie chart for the percentage distribution.
■ APPLICATIONS
2.5 The data on the status of 50 students given in Table 2.2 of Section 2. 1 are reproduced here.
J
F
SO
SE
J
J
SE
J
J
J
F
F
J
F
F
F
SE
SO
SE
J
J
F
SE
SO
SO
F
J
F
SE
SE
SO
SE
J
SO
SO
J
J
SO
F
SO
SE
SE
F
SE
J
SO
F
J
SO
SO
a. Prepare a frequency distribution table.
b. Calculate the relative frequencies and percentages for all categories.
c. What percentage of these students are juniors or seniors?
d. Draw a bar graph for the frequency distribution.
34 Chapter 2 Organizing and Graphing Data
2.6 Thirty adults were asked which of the following conveniences they would find most difficult to do
without: television (T), refrigerator (R), air conditioning (A), public transportation (P), or microwave (M).
Their responses are listed below.
R
A
R
P
P
T
R
M
P
A
A
R
R
T
P
P
T
R
A
A
R
P
A
T
R
P
R
A
P
R
a. Prepare a frequency distribution table.
b. Calculate the relative frequencies and percentages for all categories.
c. What percentage of these adults named refrigerator or air conditioning as the convenience that they
would find most difficult to do without?
d. Draw a bar graph for the relative frequency distribution.
2.7 In a USA TODAY survey, registered dietitians with the American Dietetic Association were asked,
"What is the major reason people want to lose weight?" The responses were classified as Health (H),
Cosmetic (C), and Other (O). Suppose a random sample of 20 dietitians is taken and these dietitians are
asked the same question. Their responses are as follows.
HHCHOCCHCO
OHCHHCHH OH
a. Prepare a frequency distribution table.
b. Compute the relative frequencies and percentages for all categories.
c. What percentage of these dietitians gave Health as the major reason for people to lose weight?
d. Draw a pie chart for the percentage distribution.
2.8 The following data show the method of payment by 16 customers in a supermarket checkout line.
Here, C refers to cash, CK to check, CC to credit card, and D to debit card, and O stands for other.
C CK CK C CC D O C
CK CC D CC C CK CK CC
a. Construct a frequency distribution table.
b. Calculate the relative frequencies and percentages for all categories.
c. Draw a pie chart for the percentage distribution.
2.9 In a lanuary 27, 2009 Harris Poll (Harris Interactive Inc, January 2009), U.S. adults who follow at
least one sport were asked to name their favorite sport. The table below summarizes their responses.
Favorite Sport
Percentage of Responses
Pro football
31
Baseball
16
College football
12
Auto racing
8
Men's pro basketball
6
Hockey
5
Men's college basketball
5
Note that these percentages add up to 83%. The remaining respondents named other sports, which can be
denoted by Other. Draw a pie chart for this distribution.
2.10 In exit polls taken during the 2008 presidential election, voters were asked to provide their educa-
tion levels. The table below summarizes their responses.
Education Level
Percentage of Responses
Not a high school graduate
4
High school graduate
20
Some college education
31
College graduate
28
Post graduate education
17
Source: New York Times, November 5, 2008.
Draw a bar graph to display these data.
2.3 Organizing and Graphing Quantitative Data
35
2.3 Organizing and Graphing Quantitative Data
In the previous section we learned how to group and display qualitative data. This section ex-
plains how to group and display quantitative data.
2.3.1 Frequency Distributions
Table 2.7 gives the weekly earnings of 100 employees of a large company. The first column
lists the classes, which represent the (quantitative) variable weekly earnings. For quantitative
data, an interval that includes all the values that fall within two numbers — the lower and upper
limits — is called a class. Note that the classes always represent a variable. As we can observe,
the classes are nonoverlapping; that is, each value on earnings belongs to one and only one
class. The second column in the table lists the number of employees who have earnings within
each class. For example, 9 employees of this company earn $801 to $1000 per week. The num-
bers listed in the second column are called the frequencies, which give the number of values
that belong to different classes. The frequencies are denoted by f.
Variable
Third class
Table 2.7 Weekly Earnings of 100 Employees
of a Company
Lower limit of
the sixth class
Weekly Earnings
(dollars)
Number of Employees
/
801 to 1000
9
1001 to 1200
22
1 1201 to 1400 1
m < —
1401 to 1600
15
1601 to 1800
9
|l80l|to|2000h
6
Upper limit of
Frequency column
J Frequency of
1 the third class
the sixth class
For quantitative data, the frequency of a class represents the number of values in the data
set that fall in that class. Table 2.7 contains six classes. Each class has a lower limit and an
upper limit. The values 801, 1001, 1201, 1401, 1601, and 1801 give the lower limits, and the
values 1000, 1200, 1400, 1600, 1800, and 2000 are the upper limits of the six classes, respec-
tively. The data presented in Table 2.7 are an illustration of a frequency distribution table for
quantitative data. Whereas the data that list individual values are called ungrouped data, the data
presented in a frequency distribution table are called grouped data.
Definition
Frequency Distribution for Quantitative Data A frequency distribution for quantitative data lists
all the classes and the number of values that belong to each class. Data presented in the form of
a frequency distribution are called grouped data.
To find the midpoint of the upper limit of the first class and the lower limit of the second
class in Table 2.7, we divide the sum of these two limits by 2. Thus, this midpoint is
1000 + 1001
= 1000.5
2
The value 1000.5 is called the upper boundary of the first class and the lower boundary of the
second class. By using this technique, we can convert the class limits of Table 2.7 to class
boundaries, which are also called real class limits. The second column of Table 2.8 lists the
boundaries for Table 2.7.
36 Chapter 2 Organizing and Graphing Data
Definition
Class Boundary The class boundary is given by the midpoint of the upper limit of one class and
the lower limit of the next class.
The difference between the two boundaries of a class gives the class width. The class width
is also called the class size.
Finding Class Width
Class width = Upper boundary — Lower boundary
Thus, in Table 2.8,
Width of the first class = 1000.5 - 800.5 = 200
The class widths for the frequency distribution of Table 2.7 are listed in the third column of
Table 2.8. Each class in Table 2.8 (and Table 2.7) has the same width of 200.
The class midpoint or mark is obtained by dividing the sum of the two limits (or the two
boundaries) of a class by 2.
Calculating Class Midpoint or Mark
Lower limit + Upper limit
Class midpoint or mark =
Thus, the midpoint of the first class in Table 2.7 or Table 2.8 is calculated as follows:
801 + 1000
Midpoint of the first class = = 900.5
The class midpoints for the frequency distribution of Table 2.7 are listed in the fourth column
of Table 2.8.
Table 2.8 Class Boundaries, Class Widths, and Class Midpoints for Table 2.7
Class Limits
Class Boundaries
Class Width
Class Midpoint
801 to 1000
800.5 to less than 1000.5
200
900.5
1001 to 1200
1000.5 to less than 1200.5
200
1100.5
1201 to 1400
1200.5 to less than 1400.5
200
1300.5
1401 to 1600
1400.5 to less than 1600.5
200
1500.5
1601 to 1800
1600.5 to less than 1800.5
200
1700.5
1801 to 2000
1800.5 to less than 2000.5
200
1900.5
Note that in Table 2.8, when we write classes using class boundaries, we write to less
than to ensure that each value belongs to one and only one class. As we can see, the upper
boundary of the preceding class and the lower boundary of the succeeding class are the same.
2.3.2 Constructing Frequency Distribution Tables
When constructing a frequency distribution table, we need to make the following three major
decisions.
2.3 Organizing and Graphing Quantitative Data 37
Number of Classes
Usually the number of classes for a frequency distribution table varies from 5 to 20, depending
mainly on the number of observations in the data set. 1 It is preferable to have more classes as
the size of a data set increases. The decision about the number of classes is arbitrarily made by
the data organizer.
Class Width
Although it is not uncommon to have classes of different sizes, most of the time it is preferable
to have the same width for all classes. To determine the class width when all classes are the same
size, first find the difference between the largest and the smallest values in the data. Then, the ap-
proximate width of a class is obtained by dividing this difference by the number of desired classes.
Calculation of Class Width
Largest value — Smallest value
Approximate class width
Number of classes
Usually this approximate class width is rounded to a convenient number, which is then used
as the class width. Note that rounding this number may slightly change the number of classes
initially intended.
Lower Limit of the First Class or the Starting Point
Any convenient number that is equal to or less than the smallest value in the data set can be
used as the lower limit of the first class.
Example 2-3 illustrates the procedure for constructing a frequency distribution table for
quantitative data.
■ EXAMPLE 2-3
The following data give the total number of iPods® sold by a mail order company on each of
30 days. Construct a frequency distribution table.
8
25
11
15
29
22
10
5
17
21
22
13
26
16
18
12
9
26
20
16
23
14
19
23
20
16
27
16
21
14
Solution In these data, the minimum value is 5, and the maximum value is 29. Suppose we
decide to group these data using five classes of equal width. Then,
29-5
Approximate width of each class = — - — = 4.8
Now we round this approximate width to a convenient number, say 5. The lower limit of the
first class can be taken as 5 or any number less than 5. Suppose we take 5 as the lower limit
of the first class. Then our classes will be
5-9, 10-14, 15-19, 20-24, and 25-29
We record these five classes in the first column of Table 2.9.
Constructing a
frequency distribution table for
quantitative data.
'One rale to help decide on the number of classes is Sturge's formula:
c = 1 + 3.3 log n
where c is the number of classes and n is the number of observations in the data set. The value of log n can be ob-
tained by using a calculator.
38 Chapter 2 Organizing and Graphing Data
Now we read each value from the given data and mark a tally in the second column of
Table 2.9 next to the corresponding class. The first value in our original data set is 8, which be-
longs to the 5-9 class. To record it, we mark a tally in the second column next to the 5-9 class.
We continue this process until all the data values have been read and entered in the tally column.
Note that tallies are marked in blocks of five for counting convenience. After the tally column is
completed, we count the tally marks for each class and write those numbers in the third column.
This gives the column of frequencies. These frequencies represent the number of days on which
iPods indicated in classes are sold. For example, on 8 of 30 days, 15 to 19 iPods were sold.
Table 2.9 Frequency Distribution for the Data
on iPods Sold
iPods Sold
Tally
/
5-9
3
10-14
mi
6
15-19
mm
8
20-24
mm
8
25-29
m
5
Xf= 30
In Table 2.9, we can denote the frequencies of the five classes by f lt / 2 , f$, fa, and/ 5 , respec-
tively. Therefore,
fi = Frequency of the first class = 3
Similarly,
f 2 = 6, / 3 = 8, / 4 = 8, and f 5 = 5
Using the 2 notation (see Section 1.8 of Chapter 1), we can denote the sum of frequencies of
all classes by 2/. Hence,
2/ = /, +/ 2 +f 3 + fa +f 5 = 3 + 6 + 8 + 8 + 5 = 30
The number of observations in a sample is usually denoted by n. Thus, for the sample data,
2/ is equal to n. The number of observations in a population is denoted by N. Consequently,
2/ is equal to N for population data. Because the data set on the total iPods sold on 30 days
in Table 2.9 is for only 30 days, it represents a sample. Therefore, in Table 2.9 we can denote
the sum of frequencies by n instead of 2/. B
Note that when we present the data in the form of a frequency distribution table, as in
Table 2.9, we lose the information on individual observations. We cannot know the exact num-
bers of iPods sold on any given day from Table 2.9. All we know is that for 3 days, 5 to 9 iPods
were sold, and so forth.
2.3.3 Relative Frequency and Percentage Distributions
Using Table 2.10, we can compute the relative frequency and percentage distributions in the
same way as we did for qualitative data in Section 2.2.2. The relative frequencies and percent-
ages for a quantitative data set are obtained as follows.
Calculating Relative Frequency and Percentage
Frequency of that class /
Relative frequency of a class = = -— ;
Sum of all frequencies 2 /
Percentage = (Relative frequency) • 100
Example 2-4 illustrates how to construct relative frequency and percentage distributions.
2.3 Organizing and Graphing Quantitative Data 39
■ EXAMPLE 2-4
Calculate the relative frequencies and percentages for Table 2.9.
Solution The relative frequencies and percentages for the data in Table 2.9 are calculated
and listed in the third and fourth columns, respectively, of Table 2.10. Note that the class
boundaries are listed in the second column of Table 2.10.
Table 2.10 Relative Frequency and Percentage Distributions for Table 2.9
iPods Sold
Class Boundaries
Relative Frequency
Percentage
5-9
4.5 to less than 9.5
3/30 = .100
10.0
10-14
9.5 to less than 14.5
6/30 = .200
20.0
15-19
14.5 to less than 19.5
8/30 = .267
26.7
20-24
19.5 to less than 24.5
8/30 = .267
26.7
25-29
24.5 to less than 29.5
5/30 = .167
16.7
Sum = 1.001
Sum = 100.1
Using Table 2.10, we can make statements about the percentage of days with iPods sold
within a certain interval. For example, on 20% of the days, 10 to 14 iPods were sold. By adding
the percentages for the first two classes, we can state that 5 to 14 iPods were sold on 30% of
the days. Similarly, by adding the percentages of the last two classes, we can state that 20 to
29 iPods were sold on 43.4% of the days. H
2.3.4 Graphing Grouped Data
Grouped (quantitative) data can be displayed in a histogram or a polygon. This section de-
scribes how to construct such graphs. We can also draw a pie chart to display the percentage
distribution for a quantitative data set. The procedure to construct a pie chart is similar to the
one for qualitative data explained in Section 2.2.3; it will not be repeated in this section.
Histograms
A histogram can be drawn for a frequency distribution, a relative frequency distribution, or
a percentage distribution. To draw a histogram, we first mark classes on the horizontal axis
and frequencies (or relative frequencies or percentages) on the vertical axis. Next, we draw
a bar for each class so that its height represents the frequency of that class. The bars in a
histogram are drawn adjacent to each other with no gap between them. A histogram is called
a frequency histogram, a relative frequency histogram, or a percentage histogram
depending on whether frequencies, relative frequencies, or percentages are marked on the
vertical axis.
Definition
Histogram A histogram is a graph in which classes are marked on the horizontal axis and the
frequencies, relative frequencies, or percentages are marked on the vertical axis. The frequen-
cies, relative frequencies, or percentages are represented by the heights of the bars. In a his-
togram, the bars are drawn adjacent to each other.
Figures 2.3 and 2.4 show the frequency and the relative frequency histograms, respectively,
for the data of Tables 2.9 and 2.10 of Sections 2.3.2 and 2.3.3. The two histograms look alike
because they represent the same data. A percentage histogram can be drawn for the percentage
distribution of Table 2.10 by marking the percentages on the vertical axis.
Constructing relative
frequency and percentage
distributions.
MORNING
GROOMING
USA TODAY Snapshots®
Morning grooming
How much time
do you spend on
hygiene/grooming in
the morning (including
showering, washing face
and hands, brushing
teeth, shaving, applying
makeup)?
16%
53%
5% J|
4%
Mi
0-5 6-10 11-30 31-60 More
than
60
sou tee: sca "Hygiene Miners', survey of MS 3 adults
In minutes
By Michelle Hut; and Web Bryant. USA TODaV
Source: USA Today, February 24, 2009.
Copyright © 2009, USA Today. Chart
reproduced with permission.
The accompanying chart gives the bar chart for the percentage distribution of time that adults spend on hy-
giene/grooming (such as showering, washing face and hands, brushing teeth, shaving, applying makeup) in the
morning. The results are based on a survey of 1453 adults. For example, 5% of adults included in the sample
said that they spend to 5 minutes on such activities in the morning. Note that the percentages add up to 1 01 %
due to rounding. Also note that all the classes have different widths. The last class (more than 60) is called an
open-ended class. We know that it has a lower limit of more than 60, but it does not have an upper limit.
In Figures 2.3 and 2.4, we have used class limits to mark classes on the horizontal axis.
However, we can show the intervals on the horizontal axis by using the class boundaries instead
of the class limits.
|6
£ 4
HI-
□
5-9 10-14 15-19 20-24 25-29
iPods sold
Figure 2.3 Frequency histogram for Fable 2.9.
.30
>.
o
c
a>
= .20
0)
a> .10
DC
Hh
5-9 10-14 15-19 20-24 25-29
iPods sold
Figure 2.4 Relative frequency histogram
for Fable 2.10.
Polygons
A polygon is another device that can be used to present quantitative data in graphic form.
Fo draw a frequency polygon, we first mark a dot above the midpoint of each class at a
height equal to the frequency of that class. Fhis is the same as marking the midpoint at the
40
2.3 Organizing and Graphing Quantitative Data 41
top of each bar in a histogram. Next we mark two more classes, one at each end, and mark
their midpoints. Note that these two classes have zero frequencies. In the last step, we join
the adjacent dots with straight lines. The resulting line graph is called a frequency polygon
or simply a polygon.
A polygon with relative frequencies marked on the vertical axis is called a relative fre-
quency polygon. Similarly, a polygon with percentages marked on the vertical axis is called a
percentage polygon.
Definition
Polygon A graph formed by joining the midpoints of the tops of successive bars in a histogram
with straight lines is called a polygon.
Figure 2.5 shows the frequency polygon for the frequency distribution of Table 2.9.
8
0)
—II—
I
5-9 10-14 15-19 20-24 25-29
Figure 2.5 Frequency polygon for Table 2.9.
For a very large data set, as the number of classes is increased (and the width of classes is
decreased), the frequency polygon eventually becomes a smooth curve. Such a curve is called
a frequency distribution curve or simply a frequency curve. Figure 2.6 shows the frequency
curve for a large data set with a large number of classes.
>.
o
c
<D
D
ST
0)
i_
Li.
X
Figure 2.6 Frequency distribution curve.
2.3.5 More on Classes and Frequency Distributions
This section presents two alternative methods for writing classes to construct a frequency dis-
tribution for quantitative data.
Less-Than Method for Writing Classes
The classes in the frequency distribution given in Table 2.9 for the data on iPods sold were writ-
ten as 5-9, 10-14, and so on. Alternatively, we can write the classes in a frequency distribution
table using the less-than method. The technique for writing classes shown in Table 2.9 is more
42 Chapter 2 Organizing and Graphing Data
commonly used for data sets that do not contain fractional values. The less-than method is more
appropriate when a data set contains fractional values. Example 2-5 illustrates the less-than
method.
■ EXAMPLE 2-5
On April 1, 2009, the federal tax on a pack of cigarettes was increased from 390 to $1.0066,
a move that not only was expected to help increase federal revenue, but was also expected to
save about 900,000 lives (Time Magazine, April 2009). Table 2.11 shows the total tax (state
plus federal) per pack of cigarettes for all 50 states as of April 1, 2009.
Table 2.11 Total Tax per Pack of Cigarettes
Total Tax
Total Tax
State*
(in rlnlltir^i
State
AL
1.43
MT
2.71
AK
3.01
NE
1.65
AZ
3.01
NV
1.81
AR
2.16
NH
2.34
CA
1.88
NJ
3.58
CO
1.85
NM
1.92
CT
3.01
NY
3.76
DE
2.16
NC
1.36
FL
1.35
ND
1.45
GA
1.38
OH
2.26
HI
3.01
OK
2.04
ID
1.58
OR
2.19
IL
1.99
PA
2.36
IN
2.00
RI
3.47
IA
2.37
SC
1.08
KS
1.80
SD
2.54
KY
1.61
TN
1.63
LA
1.37
TX
2.42
ME
3.01
UT
1.70
MD
3.01
VT
3.00
MA
3.52
VA
1.31
MI
3.01
WA
3.03
MN
2.51
WV
1.56
MS
1.19
WI
2.78
MO
1.18
WY
1.61
Source:
Campaign for Tobacco-Free Kids.
Construct a frequency distribution table. Calculate the relative frequencies and percentages for
all classes.
Solution The minimum value in this data set on cigarette taxes given in Table 2.11 is 1.08
and the maximum value is 3.76. Suppose we decide to group these data using six classes of
equal width. Then
3.76 - 1.08
Approximate width of a class = = .45
Constructing a frequency
distribution using the
less-than method.
2.3 Organizing and Graphing Quantitative Data 43
We round this number to a more convenient number, say .50. Then we take .50 as the width
of each class. We can take a lower limit of the first class equal to 1.08 or any number lower
than 1.08. If we start the first class at 1, the classes will be written as 1 to less than 1.5, 1.5
to less than 2.00, and so on. The six classes, which cover all the data values, are recorded in
the first column of Table 2.12. The second column lists the frequencies of these classes. A
value in the data set that is 1 or larger but less than 1 .5 belongs to the first class, a value that
is 1.50 or larger but less than 2.00 falls in the second class, and so on. The relative frequen-
cies and percentages for classes are recorded in the third and fourth columns, respectively, of
Table 2.12. Note that this table does not contain a column of tallies.
Table 2.12 Frequency, Relative Frequency, and Percentage Distributions of the Total
Tax on a Pack of Cigarettes
Total Tax (in dollars)
/
Relative
Frequency
Percentage
1.00 to less than 1.50
10
.20
20
1.50 to less than 2.00
13
.26
26
2.00 to less than 2.50
10
.20
20
2.50 to less than 3.00
4
.08
8
3.00 to less than 3.50
10
.20
20
3.50 to less than 4.00
3
.06
6
tf= 50
Sum = 1.00
Sum = 100
A histogram and a polygon for the data of Table 2.12 can be drawn the same way as for
the data of Tables 2.9 and 2.10.
Single-Valued Classes
If the observations in a data set assume only a few distinct (integer) values, it may be ap-
propriate to prepare a frequency distribution table using single-valued classes — that is,
classes that are made of single values and not of intervals. This technique is especially use-
ful in cases of discrete data with only a few possible values. Example 2-6 exhibits such a
situation.
■ EXAMPLE 2-6
The administration in a large city wanted to know the distribution of vehicles owned by house-
holds in that city. A sample of 40 randomly selected households from this city produced the
following data on the number of vehicles owned.
5
1
2
4
2
2
1
2
1
2
2
2
1
4
1
4
1
3
Construct a frequency distribution table for these data using single-valued classes.
Solution The observations in this data set assume only six distinct values: 0, 1, 2, 3, 4,
and 5. Each of these six values is used as a class in the frequency distribution in Table 2.13,
and these six classes are listed in the first column of that table. To obtain the frequencies of
these classes, the observations in the data that belong to each class are counted, and the results
are recorded in the second column of Table 2.13. Thus, in these data, 2 households own no
vehicle, 18 own one vehicle each, 11 own two vehicles each, and so on.
Constructing a
frequency distribution using
single-valued classes.
44
Chapter 2 Organizing and Graphing Data
Table 2.13 Frequency Distribution of
Vehicles Owned
Number of
Vehicles Owned
Households (/)
2
1
18
2
11
3
4
4
3
5
2
tf =40
The data of Table 2.13 can also be displayed in a bar graph, as shown in Figure 2.7. To
construct a bar graph, we mark the classes, as intervals, on the horizontal axis with a little
gap between consecutive intervals. The bars represent the frequencies of respective classes.
The frequencies of Table 2.13 can be converted to relative frequencies and percentages the
same way as in Table 2.11. Then, a bar graph can be constructed to display the relative frequency
or percentage distribution by marking the relative frequencies or percentages, respectively, on
the vertical axis.
18
15
c 1
3
o-
£ 9
LL
12 3 4
Vehicles owned
Figure 2.7 Bar graph for Table 2.13.
2.4 Shapes of Histograms
A histogram can assume any one of a large number of shapes. The most common of these
shapes are
1. Symmetric
2. Skewed
3. Uniform or rectangular
A symmetric histogram is identical on both sides of its central point. The histograms
shown in Figure 2.8 are symmetric around the dashed lines that represent their central points.
2.4 Shapes of Histograms 45
Variable
Figure 2.8 Symmetric histograms.
Variable
A skewed histogram is nonsymmetric. For a skewed histogram, the tail on one side is
longer than the tail on the other side. A skewed-to-the-right histogram has a longer tail on the
right side (see Figure 2.9a). A skewed-to-the-left histogram has a longer tail on the left side
(see Figure 2.9b).
Variable Variable
(a) (b)
Figure 2.9 (a) A histogram skewed to the right, (b) A histogram skewed to the left.
A uniform or rectangular histogram has the same frequency for each class. Figure 2.10
is an illustration of such a case.
>.
u
c
CD
3
!T
CD
Figure 2. TO A histogram with uniform
distribution.
Variable
Figures 2.1 la and 2.1 lb display symmetric frequency curves. Figures 2.1 lc and 2.1 la" show
frequency curves skewed to the right and to the left, respectively.
Figure 2.1 1 (a), (b) Symmetric frequency curves, (c) Frequency curve skewed to
the right, (d) Frequency curve skewed to the left.
46 Chapter 2 Organizing and Graphing Data
Warning ► Describing data using graphs give us insights into the main characteristics of the data. But
graphs, unfortunately, can also be used, intentionally or unintentionally, to distort the facts and
deceive the reader. The following are two ways to manipulate graphs to convey a particular
opinion or impression.
1. Changing the scale either on one or on both axes — that is, shortening or stretching one or
both of the axes.
2. Truncating the frequency axis — that is, starting the frequency axis at a number greater than zero.
When interpreting a graph, we should be very cautious. We should observe carefully whether
the frequency axis has been truncated or whether any axis has been unnecessarily shortened or
stretched. See the Uses and Misuses section of this chapter for such an example.
I
EXERCISES
CONCEPTS AND PROCEDURES
2.11 Briefly explain the three decisions that have to be made to group a data set in the form of a fre-
quency distribution table.
2.12 How are the relative frequencies and percentages of classes obtained from the frequencies of classes?
Illustrate with the help of an example.
2.13 Three methods — writing classes using limits, using the less-than method, and grouping data using
single-valued classes — were discussed to group quantitative data into classes. Explain these three meth-
ods and give one example of each.
■ APPLICATIONS
2.14 A sample of 80 adults was taken, and these adults were asked about the number of credit cards they
possess. The following table gives the frequency distribution of their responses.
Number of Credit Cards Number of Adults
to 3 18
4 to 7 26
8 to 1 1 22
12 to 15 11
16 to 19 3
a. Find the class boundaries and class midpoints.
b. Do all classes have the same width? If so, what is this width?
c. Prepare the relative frequency and percentage distribution columns.
d. What percentage of these adults possess 8 or more credit cards?
2.15 The following table gives the frequency distribution of ages for all 50 employees of a company.
Age
Number of Employees
18 to 30
12
31 to 43
19
44 to 56
14
57 to 69
5
a. Find the class boundaries and class midpoints.
b. Do all classes have the same width? If yes, what is that width?
c. Prepare the relative frequency and percentage distribution columns.
d. What percentage of the employees of this company are age 43 or younger?
2.16 A data set on money spent on lottery tickets during the past year by 200 households has a lowest value
of $1 and a highest value of SI 167. Suppose we want to group these data into six classes of equal widths.
a. Assuming that we take the lower limit of the first class as $1 and the width of each class equal
to S200, write the class limits for all six classes.
b. What are the class boundaries and class midpoints?
2.4 Shapes of Histograms 47
2.17 A data set on monthly expenditures (rounded to the nearest dollar) incurred on fast food by a sam-
ple of 500 households has a minimum value of S3 and a maximum value of $147. Suppose we want to
group these data into six classes of equal widths.
a. Assuming that we take the lower limit of the first class as $1 and the upper limit of the sixth class
as $150, write the class limits for all six classes.
b. Determine the class boundaries and class widths.
c. Find the class midpoints.
2.18 The accompanying table lists the 2006-07 median household incomes (rounded to the nearest dollar),
for all 50 states and the District of Columbia.
State
'yiUif^ftl Mori! -in
zuuo— u / ivieuian
Household Income
State
zuuo— u / ivieuian
Household Income
AL
40,620
MT
42,963
AK
60,506
NE
49,342
AZ
47,598
NV
53,912
AR
39,452
NH
65,652
CA
56,311
NJ
65,249
CO
59,209
NM
42,760
CT
64,158
NY
49,267
DE
54,257
NC
42,219
D.C.
50,318
ND
44,708
FL
46,383
OH
48,151
GA
49,692
OK
41,578
HI
63,104
OR
49,331
ID
48,354
PA
49,145
IL
51,279
RI
54,735
IN
47,074
SC
42,477
IA
49,200
SD
46,567
KS
47,671
TN
41,521
KY
40,029
TX
45,294
LA
39,418
UT
54,853
ME
47,415
VT
50,423
MD
65,552
VA
58,950
MA
57,681
WA
57,178
MI
49,699
WV
40,800
MN
57,932
WI
52,218
MS
36,499
WY
48,560
MO
45,924
Source: U.S. Census Bureau.
a. Construct a frequency distribution table. Use the following classes: 36,000^-0,999, 41,000-
45,999, 46,000-50,999, 51,000-55,999, 56,000-60,999, 61,000-65,999.
b. Calculate the relative frequencies and percentages for all classes.
c. Based on the frequency distribution, can you say whether the data are symmetric or skewed?
d. What percentage of these states had a median household income of less than $56,000?
2.19 Nixon Corporation manufactures computer monitors. The following data are the numbers of com-
puter monitors produced at the company for a sample of 30 days.
24
32
27
23
33
33
29
25
23
28
21
26
31
22
27
33
27
23
28
29
31
35
34
22
26
28
23
35
31
27
48 Chapter 2 Organizing and Graphing Data
a. Construct a frequency distribution table using the classes 21-23, 24-26, 27-29, 30-32, and 33-35.
b. Calculate the relative frequencies and percentages for all classes.
c. Construct a histogram and a polygon for the percentage distribution.
d. For what percentage of the days is the number of computer monitors produced in the interval
27-29?
2.20 The following data give the numbers of computer keyboards assembled at the Twentieth Century
Electronics Company for a sample of 25 days.
45 52 48 41 56 46 44 42 48 53 51 53 51
48 46 43 52 50 54 47 44 47 50 49 52
a. Make the frequency distribution table for these data.
b. Calculate the relative frequencies for all classes.
c. Construct a histogram for the relative frequency distribution.
d. Construct a polygon for the relative frequency distribution.
2.21 Since 1996, Slate.com has determined the Slate 60, which is a list of the largest American charita-
ble contributions by individuals each year. The accompanying table gives the names of the 22 persons and
the money they donated in 2008.
Donation
Donor
(millions of dollars)
Harold Alfond
360
Donald B. and Dorothy L. Stabler
334.2
David G. and Suzanne D. Booth
300
Frank C. Doble
272
Robert L. and Catherine H. McDevitt
250
Michael R. Bloomberg
235
Dorothy Clarke Patterson
225
Richard W. Weiland
174.3
Helen L. Kimmel
156.5
Jeffrey S. Skoll
144.1
H. F. (Gerry) and Marguerite B. Lenfest
139.9
David Rockefeller
137.8
Stephen A. Schwarzman
105
David H. Koch
LOO
Gerhard R. Andlinger
100
Eli and Edythe L. Broad
100
Philip H. and Penelope Knight
100
Kenneth G. and Elaine A. Langone
100
Fritz J. and Dolores H. Russ
94.8
Frank Sr. and Jane Batten
93
Jesse H. and Beulah C. Cox
83.5
Henry R. and Marie-Josee Kravis
75
Source: Slate.com, January 26, 2009.
a. Construct a frequency distribution table using the following classes: 75 to less than 125, 125 to
less than 175, 175 to less than 225, and so on.
b. Calculate the relative frequencies and percentages for all classes.
Exercises 2.22 through 2.26 are based on the following data.
The following table gives the age-adjusted cancer incidence rates (new cases) per 100,000 people for
three of the most common types of cancer contracted by both females and males: colon and rectum
2.4 Shapes of Histograms 49
cancer, lung and bronchus cancer, and non-hodgkin lymphoma. The rates given are for 22 states west
of the Mississippi River for the years 2000 to 2004 (except for South Dakota, for which the data are
for 2001 to 2004), which are the most recent data available from the American Cancer Society. Age-
adjusted rates take into account the percentage of people in different age groups within each state's
population.
State
Colon and
Rectum
(Males)
Colon and
Rectum
(Females)
Lung and
Bronchus
(Males)
Lung and
Bronchus
(Females)
Non-Hodgkin
Lymphoma
(Males)
Non-Hodgkin
Lymphoma
(Females)
AK
63.9
50.0
87.0
59.2
24.0
14.6
AZ
52.3
37.4
71.2
49.3
18.9
13.4
AR
60.5
43.1
113.7
57.8
20.9
15.2
CA
55.0
40.4
69.0
47.9
22.4
15.4
CO
53.1
41.4
65.1
45.9
21.6
16.6
HI
64.7
41.6
67.8
37.6
17.9
13.2
ID
51.9
39.7
71.4
45.8
21.4
17.7
1A
69.0
51.5
89.5
50.8
22.8
17.1
LA
71.3
48.9
112.3
57.4
22.8
16.0
MO
65.9
47.5
104.3
60.0
22.0
16.0
MT
56.5
43.3
79.8
56.7
22.8
15.0
NE
69.1
48.6
84.0
47.8
23.8
17.2
NV
59.8
44.6
88.8
71.4
22.0
15.5
NM
51.5
35.7
59.5
37.7
18.1
13.9
ND
66.3
43.3
71.3
43.9
22.1
15.1
OK
63.1
44.6
109.2
63.1
22.0
15.7
OR
56.6
42.6
81.4
61.0
24.2
17.3
SD
66.4
47.5
82.0
43.3
22.5
17.1
TX
59.7
41.0
91.0
51.0
21.7
15.9
UT
47.5
35.2
40.3
20.9
23.2
15.7
WA
55.9
42.5
82.0
60.5
26.6
18.3
WY
49.5
43.4
65.3
44.8
18.9
17.6
Source: American Cancer Society, 2008.
2.22 a. Prepare a frequency distribution table for colon and rectum cancer rates for women using six
classes of equal width,
b. Construct the relative frequency and percentage distribution columns.
2.23 a. Prepare a frequency distribution table for colon and rectum cancer rates for men using six classes
of equal width.
b. Construct the relative frequency and percentage distribution columns.
2.24 a. Prepare a frequency distribution table for lung and bronchus cancer rates for women.
b. Construct the relative frequency and percentage distribution columns.
c. Draw a histogram and polygon for the relative frequency distribution.
2.25 a. Prepare a frequency distribution table for lung and bronchus cancer rates for men.
b. Construct the relative frequency and percentage distribution columns.
c. Draw a histogram and polygon for the relative frequency distribution.
2.26 a. Prepare a frequency distribution table for non-Hodgkin lymphoma rates for women.
b. Construct the relative frequency and percentage distribution columns.
c. Draw a histogram and polygon for the relative frequency distribution.
2.27 The accompanying table lists the offensive points scored per game (PPG) by each of the 16 teams in
the American Football Conference (AFC) of the National Football League (NFL) during the 2008 season.
50 Chapter 2 Organizing and Graphing Data
Team
PPG
Team
PPG
Baltimore
24.1
Kansas City
18.2
Buffalo
21.0
Miami
21.6
Cincinnati
12.8
New England
25.6
Cleveland
14.5
New York Jets
25.3
Denver
23.1
Oakland
16.4
Houston
22.9
Pittsburgh
21.7
Indianapolis
23.6
San Diego
27.4
Jacksonville
18.9
Tennessee
23.4
a. Construct a frequency distribution table. Take 12.0 as the lower boundary of the first class and
3.5 as the width of each class.
b. Prepare the relative frequency and percentage distribution columns for the frequency table of part a.
2.28 The following data give the number of turnovers (fumbles and interceptions) by a college football
team for each game in the past two seasons.
321402210323
023141324012
a. Prepare a frequency distribution table for these data using single- valued classes.
b. Calculate the relative frequencies and percentages for all classes.
c. In how many games did the team commit two or more turnovers?
d. Draw a bar graph for the frequency distribution of part a.
2.29 According to a survey by the U.S. Public Interest Research Group, about 79% of credit reports contain
errors. Suppose in a random sample of 25 credit reports, the number of errors found are as listed below.
1023010541021
412203100123
a. Prepare a frequency distribution table for these data using single- valued classes.
b. Calculate the relative frequencies and percentages for all classes.
c. How many of these reports contained two or more errors?
d. Draw a bar graph for the frequency distribution of part a.
2.30 The following table gives the frequency distribution for the numbers of parking tickets received on
the campus of a university during the past week for 200 students.
Number of Tickets
Number of Students
59
1
44
2
37
3
32
4
28
Draw two bar graphs for these data, the first without truncating the frequency axis and the second by trun-
cating the frequency axis. In the second case, mark the frequencies on the vertical axis starting with 25.
Briefly comment on the two bar graphs.
2.31 Eighty adults were asked to watch a 30-minute infomercial until the presentation ended or until bore-
dom became intolerable. The following table lists the frequency distribution of the times that these adults
were able to watch the infomercial.
Time
Number of
(minutes)
Adults
to less than 6
16
6 to less than 12
21
12 to less than 18
18
18 to less than 24
11
24 to less than 30
14
2.5 Cumulative Frequency Distributions 51
Draw two histograms for these data, the first without truncating the frequency axis. In the second case,
mark the frequencies on the vertical axis starting with 10. Briefly comment on the two histograms.
2.5 Cumulative Frequency Distributions
Consider again Example 2-3 of Section 2.3.2 about the total number of iPods sold by a com-
pany. Suppose we want to know on how many days the company sold 19 or fewer iPods. Such
a question can be answered by using a cumulative frequency distribution. Each class in a cu-
mulative frequency distribution table gives the total number of values that fall below a certain
value. A cumulative frequency distribution is constructed for quantitative data only.
Definition
Cumulative Frequency Distribution A cumulative frequency distribution gives the total number
of values that fall below the upper boundary of each class.
In a cumulative frequency distribution table, each class has the same lower limit but a dif-
ferent upper limit. Example 2-7 illustrates the procedure for preparing a cumulative frequency
distribution.
■ EXAMPLE 2-7
Using the frequency distribution of Table 2.9, reproduced here, prepare a cumulative frequency
distribution for the number of iPods sold by that company.
iPods Sold /
5-9
3
10-14
6
15-19
8
20-24
8
25-29
5
Solution Table 2.14 gives the cumulative frequency distribution for the number of iPods
sold. As we can observe, 5 (which is the lower limit of the first class in Table 2.9) is taken as
the lower limit of each class in Table 2.14. The upper limits of all classes in Table 2.14 are
the same as those in Table 2.9. To obtain the cumulative frequency of a class, we add the fre-
quency of that class in Table 2.9 to the frequencies of all preceding classes. The cumulative
frequencies are recorded in the third column of Table 2.14. The second column of this table
lists the class boundaries.
Table 2.14 Cumulative Frequency Distribution of iPods Sold
Class Limits Class Boundaries Cumulative Frequency
5-9
4.5 to less than 9.5
3
5-14
4.5 to less than 14.5
3 + 6 = 9
5-19
4.5 to less than 19.5
3 + 6 + 8 = 17
5-24
4.5 to less than 24.5
3 + 6 + 8 + 8 = 25
5-29
4.5 to less than 29.5
3 + 6 + 8 + 8 + 5 = 30
From Table 2.14, we can determine the number of observations that fall below the upper
limit or boundary of each class. For example, 19 or fewer iPods were sold on 17 days. H
Constructing
a cumulative frequency
distribution table.
52 Chapter 2 Organizing and Graphing Data
The cumulative relative frequencies are obtained by dividing the cumulative frequencies
by the total number of observations in the data set. The cumulative percentages are obtained
by multiplying the cumulative relative frequencies by 100.
Calculating Cumulative Relative Frequency and Cumulative Percentage
Cumulative frequency of a class
Cumulative relative frequency =
Total observations in the data set
Cumulativep ercentage = (Cumulative relative frequency) • 100
Table 2.15 contains both the cumulative relative frequencies and the cumulative percent-
ages for Table 2.14. We can observe, for example, that 19 or fewer iPods were sold on 56.7%
of the days.
Table 2.15 Cumulative Relative Frequency and
Cumulative Percentage Distributions
for iPods Sold
Cumulative
Cumulative
Class Limits
Relative Frequency
Percentage
5-9
3/30 = .100
10.0
5-14
9/30 = .300
30.0
5-19
17/30 = .567
56.7
5-24
25/30 = .833
83.3
5-29
30/30 = 1.000
100.0
Ogives
When plotted on a diagram, the cumulative frequencies give a curve that is called an ogive
(pronounced o-jive ). Figure 2.12 gives an ogive for the cumulative frequency distribution of
Table 2.14. To draw the ogive in Figure 2.12, the variable, which is total iPods sold, is marked
on the horizontal axis and the cumulative frequencies on the vertical axis. Then the dots are
marked above the upper boundaries of various classes at the heights equal to the corresponding
cumulative frequencies. The ogive is obtained by joining consecutive points with straight lines.
Note that the ogive starts at the lower boundary of the first class and ends at the upper bound-
ary of the last class.
30
>.
o
c
<D
ZS
d)
I 15
3
E
3
o
25
20
10
— n — <
4.5 9.5
14.5 19.5
iPods sold
24.5 29.5
Figure 2.12 Ogive for the cumulative frequency
distribution of Table 2.14.
2.5 Cumulative Frequency Distributions 53
Definition
Ogive An ogive is a curve drawn for the cumulative frequency distribution by joining with
straight lines the dots marked above the upper boundaries of classes at heights equal to the cumu-
lative frequencies of respective classes.
One advantage of an ogive is that it can be used to approximate the cumulative frequency
for any interval. For example, we can use Figure 2.12 to find the number of days for which 17
or fewer iPods were sold. First, draw a vertical line from 17 on the horizontal axis up to the
ogive. Then draw a horizontal line from the point where this line intersects the ogive to the ver-
tical axis. This point gives the cumulative frequency of the class 5 to 17. In Figure 2.12, this
cumulative frequency is (approximately) 13 as shown by the dashed line. Therefore, 17 or fewer
iPods were sold on 13 days.
We can draw an ogive for cumulative relative frequency and cumulative percentage distri-
butions the same way as we did for the cumulative frequency distribution.
EXERCISES
CONCEPTS AND PROCEDURES
2.32 Briefly explain the concept of cumulative frequency distribution. How are the cumulative relative
frequencies and cumulative percentages calculated?
2.33 Explain for what kind of frequency distribution an ogive is drawn. Can you think of any use for an
ogive? Explain.
■ APPLICATIONS
2.34 The following table, reproduced from Exercise 2.14, gives the frequency distribution of the number
of credit cards possessed by 80 adults.
Number of Credit Cards Number of Adults
to 3 18
4 to 7 26
8 to 1 1 22
12 to 15 11
16 to 19 3
a. Prepare a cumulative frequency distribution.
b. Calculate the cumulative relative frequencies and cumulative percentages for all classes.
c. Find the percentage of these adults who possess 7 or fewer credit cards.
d. Draw an ogive for the cumulative percentage distribution.
e. Using the ogive, find the percentage of adults who possess 10 or fewer credit cards.
2.35 The following table, reproduced from Exercise 2.15, gives the frequency distribution of ages for all
50 employees of a company.
Age
Number of Employees
18 to 30
12
31 to 43
19
44 to 56
14
57 to 69
5
a. Prepare a cumulative frequency distribution table.
b. Calculate the cumulative relative frequencies and cumulative percentages for all classes.
54 Chapter 2 Organizing and Graphing Data
c. What percentage of the employees of this company are 44 years of age or older?
d. Draw an ogive for the cumulative percentage distribution.
e. Using the ogive, find the percentage of employees who are age 40 or younger.
2.36 Using the frequency distribution table constructed in Exercise 2.18, prepare the cumulative fre-
quency, cumulative relative frequency, and cumulative percentage distributions.
2.37 Using the frequency distribution table constructed in Exercise 2. 19, prepare the cumulative frequency,
cumulative relative frequency, and cumulative percentage distributions.
2.38 Using the frequency distribution table constructed in Exercise 2.20, prepare the cumulative frequency,
cumulative relative frequency, and cumulative percentage distributions.
2.39 Prepare the cumulative frequency, cumulative relative frequency, and cumulative percentage distributions
using the frequency distribution constructed in Exercise 2.23.
2.40 Using the frequency distribution table constructed for the data of Exercise 2.25, prepare the cumulative
frequency, cumulative relative frequency, and cumulative percentage distributions.
2.41 Refer to the frequency distribution table constructed in Exercise 2.26. Prepare the cumulative frequency,
cumulative relative frequency, and cumulative percentage distributions by using that table.
2.42 Using the frequency distribution table constructed for the data of Exercise 2.21, prepare the cumulative
frequency, cumulative relative frequency, and cumulative percentage distributions. Draw an ogive for the
cumulative frequency distribution. Using the ogive, find the (approximate) number of individuals who
made charitable contributions of $200 million or less.
2.43 Refer to the frequency distribution table constructed in Exercise 2.27. Prepare the cumulative frequency,
cumulative relative frequency, and cumulative percentage distributions. Draw an ogive for the cumulative
frequency distribution. Using the ogive, find the (approximate) number of teams with 20 or fewer offen-
sive points scored per game.
2.6 Stem-and-Leaf Displays
Another technique that is used to present quantitative data in condensed form is the stem-and-
leaf display. An advantage of a stem-and-leaf display over a frequency distribution is that by
preparing a stem-and-leaf display we do not lose information on individual observations. A
stem-and-leaf display is constructed only for quantitative data.
Definition
Stem-and-Leaf Display In a stem-and-leaf display of quantitative data, each value is divided into
two portions — a stem and a leaf. The leaves for each stem are shown separately in a display.
Example 2-8 describes the procedure for constructing a stem-and-leaf display.
Constructing a
stem-and-leaf display for
two-digit numbers.
■ EXAMPLE 2-8
The following are the scores of 30 college students on a statistics test.
75
52
80
96
65
79
71
87
93
95
69
72
81
61
76
86
79
68
50
92
83
84
77
64
71
87
72
92
57
98
Construct a stem-and-leaf display.
Solution To construct a stem-and-leaf display for these scores, we split each score into two
parts. The first part contains the first digit, which is called the stem. The second part contains
the second digit, which is called the leaf. Thus, for the score of the first student, which is 75,
7 is the stem and 5 is the leaf. For the score of the second student, which is 52, the stem is 5
and the leaf is 2. We observe from the data that the stems for all scores are 5, 6, 7, 8, and 9
because all the scores lie in the range 50 to 98. To create a stem-and-leaf display, we draw a
2.6 Stem-and-Leaf Displays 55
vertical line and write the stems on the left side of it, arranged in increasing order, as shown
in Figure 2.13.
Stems
Figure 2.13 Stem-and-leaf display.
2 < Leaf for 52
5 < Leaf for 75
After we have listed the stems, we read the leaves for all scores and record them next
to the corresponding stems on the right side of the vertical line. For example, for the first
score we write the leaf 5 next to the stem 7; for the second score we write the leaf 2 next
to the stem 5. The recording of these two scores in a stem-and-leaf display is shown in
Figure 2.13.
Now, we read all the scores and write the leaves on the right side of the vertical line in
the rows of corresponding stems. The complete stem-and-leaf display for scores is shown in
Figure 2.14.
2 7
5 9 18 4
5 9 1 2 6 9 7
7 1 6 3 4 7
6 3 5 2 2 8
Figure 2.14 Stem-and-
leaf display of test scores.
1 2
By looking at the stem-and-leaf display of Figure 2.14, we can observe how the data values
are distributed. For example, the stem 7 has the highest frequency, followed by stems 8, 9, 6,
and 5.
The leaves for each stem of the stem-and-leaf display of Figure 2. 14 are ranked (in increasing
order) and presented in Figure 2.15.
5 2 7
6 1 4 5 8 9
7 112256799
8 1 3 4 6 7 7
9 2 2 3 5 6 8
Figure 2.15 Ranked
stem-and-leaf display of
test scores.
As already mentioned, one advantage of a stem-and-leaf display is that we do not lose
information on individual observations. We can rewrite the individual scores of the 30 college
students from the stem-and-leaf display of Figure 2.14 or Figure 2.15. By contrast, the
information on individual observations is lost when data are grouped into a frequency table.
■ EXAMPLE 2-9
The following data give the monthly rents paid by a sample of 30 households selected from
a small town.
880 1081
1210 985
1151 630
721
1231
1175
1075
932
952
1023
850
1100
775
825
1140
1235
1000
750
750
915
1140
965 960
1191 1035
1370 1280
Constructing a stem-and-leaf
display for three-and four-digit
numbers.
Construct a stem-and-leaf display for these data.
56 Chapter 2 Organizing and Graphing Data
Solution Each of the values in the data set contains either three or four digits. We will take
the first digit for three-digit numbers and the first two digits for four-digit numbers as stems.
Then we will use the last two digits of each number as a leaf. Thus for the first value, which
is 880, the stem is 8 and the leaf is 80. The stems for the entire data set are 6, 7, 8, 9, 10, 11,
12, and 13. They are recorded on the left side of the vertical line in Figure 2.16. The leaves
for the numbers are recorded on the right side.
Figure 2.16 Stem-and-leaf
display of rents.
6
30
7
75
50
21
50
8
80
25
50
9
32
52
15
60
85
65
10
23
81
35
75
00
11
91
51
40
75
40
00
12
10
31
35
80
13
70
Sometimes a data set may contain too many stems, with each stem containing only a few
leaves. In such cases, we may want to condense the stem-and-leaf display by grouping the stems.
Example 2-10 describes this procedure.
EXAMPLE 2-10
Preparing a grouped
stem-and-leaf display.
The following stem-and-leaf display is prepared for the number of hours that 25 students spent
working on computers during the past month.
6
1 7 9
2 6
2 4 7 8
1 5 6 9 9
3 6 8
2 4 4 5 7
5 6
Prepare a new stem-and-leaf display by grouping the stems.
Solution To condense the given stem-and-leaf display, we can combine the first three rows,
the middle three rows, and the last three rows, thus getting the stems 0-2, 3-5, and 6-8. The
leaves for each stem of a group are separated by an asterisk (*), as shown in Figure 2. 17. Thus,
the leaf 6 in the first row corresponds to stem 0; the leaves 1, 7, and 9 correspond to stem 1;
and leaves 2 and 6 belong to stem 2.
0-26*179*26
3- 5 2478*15699*368
6-824457**56
Figure 2.17 Grouped stem-and-leaf display.
If a stem does not contain a leaf, this is indicated in the grouped stem-and-leaf display by
two consecutive asterisks. For example, in the above stem-and-leaf display, there is no leaf for
7; that is, there in no number in the 70s. Hence, in Figure 2.17, we have two asterisks after
the leaves for 6 and before the leaves for 8.
2.6 Stem-and-Leaf Displays 57
EXERCISES
CONCEPTS AND PROCEDURES
2.44 Briefly explain how to prepare a stem-and-leaf display for a data set. You may use an example to
illustrate.
2.45 What advantage does preparing a stem-and-leaf display have over grouping a data set using a fre-
quency distribution? Give one example.
2.46 Consider this stem-and-leaf display.
3 6
14 5
34677789
2 2 3 5 6 6 9
7 8 9
Write the data set that is represented by the display.
2.47 Consider this stem-and-leaf display.
2-3
4-5
6-8
18 45 56 * 29 67 83 97
04 27 33 71 * 23 37 51 63 81 92
22 36 47 55 78 89 * * 10 41
Write the data set that is represented by the display.
■ APPLICATIONS
2.48 The following data give the time (in minutes) that each of 20 students waited in line at their book-
store to pay for their textbooks in the beginning of Spring 2009 semester. {Note: To prepare a stem-and-
leaf display, each number in this data set can be written as a two-digit number. For example, 8 can be writ-
ten as 08, for which the stem is and the leaf is 8.)
15 8 23 21 5 17 31 22 34 6
5 10 14 17 16 25 30 3 31 19
Construct a stem-and-leaf display for these data. Arrange the leaves for each stem in increasing order.
2.49 Following are the total yards gained rushing during the 2009 season by 14 running backs of 14 college
football teams.
745 921 1133 1024 848 775 800
1009 1275 857 933 1145 967 995
Prepare a stem-and-leaf display. Arrange the leaves for each stem in increasing order.
2.50 Reconsider the data on the numbers of computer monitors produced at the Nixon Corporation for a
sample of 30 days given in Exercise 2.19. Prepare a stem-and-leaf display for those data. Arrange the
leaves for each stem in increasing order.
2.51 Reconsider the data on the numbers of computer keyboards assembled at the Twentieth Century Elec-
tronics Company given in Exercise 2.20. Prepare a stem-and-leaf display for those data. Arrange the leaves
for each stem in increasing order.
2.52 Refer to Exercise 2.18. Rewrite those data by rounding each median household income to the near-
est thousand. For example, a median household income of $43,260 will be rounded to 43 thousand, and
$50,689 will be rounded to 51 thousand. Prepare a stem-and-leaf display for these data. Arrange the leaves
for each stem in increasing order.
2.53 These data give the times (in minutes) taken to commute from home to work for 20 workers.
10 50 65 33 48 5 11 23 39 26
26 32 17 7 15 19 29 43 21 22
Construct a stem-and-leaf display for these data. Arrange the leaves for each stem in increasing order.
58 Chapter 2 Organizing and Graphing Data
2.54 The following data give the times served (in months) by 35 prison inmates who were released recently.
37
6
20
5
25
30
24
10
12
20
24
8
26
15
13
22
72
80
96
33
84
86
70
40
92
36
28
90
36
32
72
45
38
18
9
a. Prepare a stem-and-leaf display for these data.
b. Condense the stem-and-leaf display by grouping the stems as 0-2, 3-5, and 6-9.
2.55 The following data give the money (in dollars) spent on textbooks by 35 students during the 2009-10
academic year.
565
728
470
620
345
368
610
765
550
845
530
705
490
258
320
505
457
787
617
721
635
438
575
702
538
720
460
540
890
560
570
706
430
268
638
a. Prepare a stem-and-leaf display for these data using the last two digits as leaves.
b. Condense the stem-and-leaf display by grouping the stems as 2-4, 5-6, and 7-8.
2.7 Dotplots
One of the simplest methods for graphing and understanding quantitative data is to create a dot-
plot. As with most graphs, statistical software should be used to make a dotplot for large data
sets. However, Example 2-1 1 demonstrates how to create a dotplot by hand.
Dotplots can help us detect outliers (also called extreme values) in a data set. Outliers are
the values that are extremely large or extremely small with respect to the rest of the data values.
Definition
Outliers or Extreme Values Values that are very small or very large relative to the majority of
the values in a data set are called outliers or extreme values.
■ EXAMPLE 2-11
Table 2.16 lists the lengths of the longest field goals (in yards) made by all kickers in the
American Football Conference (AFC) of the National Football League (NFL) during the
2008 season. Create a dotplot for these data.
Table 2.16 Distances of Longest Field Goals (in Yards) Made by AFC Kickers During the
2008 NFL Season
Name
Team
Distance
Name
Team
Distance
S. Hauschka
Baltimore
54
C. Baith
Kansas City
45
M. Stover
Baltimore
47
N. Novak
Kansas City
43
R. Lindell
Buffalo
53
D. Carpenter
Miami
50
S. Graham
Cincinnati
45
S. Gostkowski
New England
50
D. Rayner
Cincinnati
26
J. Feely
New York Jets
55
P. Dawson
Cleveland
56
S. Janikowski
Oakland
57
M. Prater
Denver
56
J. Reed
Pittsburgh
53
K. Brown
Houston
53
N. Kaeding
San Diego
57
A. Vinatieri
Indianapolis
52
R. Bironas
Tennessee
51
J. Scobee
Jacksonville
53
Source: ESPN.com.
Solution Below we show how to make a dotplot for these data on field goal distances.
Step 1. The minimum and maximum values in this data set are 26 and 57 yards, respec-
tively. First, we draw a horizontal line (let us call this the numbers line) with numbers that
Creating a dotplot.
2.7 Dotplots 59
cover the given data as shown in Figure 2.18. Note that the numbers line in Figure 2.18
shows the values from 25 to 57.
— i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 — *-
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57
Figure 2.18 Numbers line.
Step 2. Place a dot above the value on the numbers line that represents each distance listed
in the table. For example, S. Hauschka's longest successful field goal of the 2008 season was
54 yards. Place a dot above 54 on the numbers line as shown in Figure 2.19. If there are two
or more observations with the same value, we stack dots vertically above each other to rep-
resent those values. For example, as shown in Table 2.16, 53 yards was the distance of the
longest field goals made by four players. We stack four dots (one for each player) above 53
on the numbers line as shown in Figure 2.19. Figure 2.19 gives the complete dotplot.
• • • • •
• ••• ••••••••
— i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 — »-
25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57
Figure 2.19 Dotplot for Longest Completed Field Goal.
As we examine the dotplot of Figure 2.19, we notice that there are two clusters (groups) of data.
Approximately 70% of the kickers made field goals of 50 to 57 yards. All but one of the re-
maining kickers completed long field goals of 43 to 47 yards. In addition, one kicker, D. Rayner
of the Cincinnati Bengals, had a longest field goal of 26 yards. When this occurs, we expect that
such a data value could be an outlier. (In the box-and-whisker section of Chapter 3, we will learn
a numerical method to determine whether a data point should be classified as an outlier.) I
Dotplots are also very useful for comparing two or more data sets. To do so, we create a
dotplot for each data set with numbers lines for all data sets on the same scale. We place these
data sets on top of each other, resulting in what are called stacked dotplots. Example 2-12
shows this procedure.
■ EXAMPLE 2-12
Refer to Table 2.16 in Example 2—11, which gives the distances of longest completed field
goals for all kickers in the AFC during the 2008 NFL season. Table 2.17 provides the same
information for the kickers in the National Football Conference (NFC) of the NFL for the 2008
season. Make dotplots for both sets of data and compare these two dotplots.
Table 2.17 Distances of Longest Field Goals (in Yards) Made by NFC Kickers During the
2008 NFL Season
Name
Team
Distance
Name
Team
Distance
N. Rackers
Arizona
54
T. Mehlhaff
New Orleans
44
J. Elam
Atlanta
50
J. Carney
New York Giants
51
J. Kasay
Carolina
50
L. Tynes
New York Giants
19
R. Gould
Chicago
48
D. Akers
Philadelphia
51
N. Folk
Dallas
52
J. Nedney
San Francisco
53
J. Hanson
Detroit
56
O. Mare
Seattle
51
M. Crosby
Green Bay
53
J. Brown
St. Louis
54
R. Longwell
Minnesota
54
M. Bryant
Tampa Bay
49
M. Gramatica
New Orleans
53
S. Suisham
Washington
50
G. Hartley
New Orleans
47
Source: ESPN.com.
Comparing two data sets
using dotplots.
60 Chapter 2 Organizing and Graphing Data
Solution Figure 2.20 shows the dotplots for the field goal distances for the kickers in both
NFL conferences.
19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57
AFC kickers
• • • •
• • • •
— I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 H
19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57
NFC kickers
Figure 2.20 Stacked dotplot of longest field goals made by kickers in the AFC and NFC during
the 2008 NFL season.
Looking at the stacked dotplot, we see that the majority of the distances fall within a range of
7 yards. In the AFC, the range is 50 to 57 yards, whereas in the NFC, the range is 47 to 54
yards. Both conferences have one outlier, with the shortest distance being 19 yards.
In practice, dotplots and other statistical graphs will be created using statistical software.
The Technology Instruction section at the end of this chapter shows how we can do so.
EXERCISES
CONCEPTS AND PROCEDURES
2.56 Briefly explain how to prepare a dotplot for a data set. You may use an example to illustrate.
2.57 What is a stacked dotplot, and how is it used? Explain.
2.58 Create a dotplot for the following data set.
■ APPLICATIONS
2.59 Reconsider the data on the numbers of computer keyboards assembled at the Twentieth Century Elec-
tronics Company given in Exercise 2.20. Create a dotplot for those data.
2.60 Create a dotplot for the data on the number of turnovers (fumbles and interceptions) by a college
football team for games in the past two seasons given in Exercise 2.28.
2.61 Reconsider the data on the numbers of errors found in 25 randomly selected credit reports given in
Exercise 2.29. Create a dotplot for those data.
2.62 The following data give the number of times each of the 30 randomly selected account holders at a
bank used that bank's ATM during a 60-day period.
3
2
3
2
2
5
4
1
3
2
3
3
5
9
3
2
2
15
1
3
2
7
9
3
4
2
2
Create a dotplot for these data and point out any clusters or outliers.
2.7 Dotplots 61
2.63 The following data give the number of times each of the 20 randomly selected male students from
a state university ate at fast-food restaurants during a 7-day period.
58 10 355 10 72 1
10 450 10 12835
Create a dotplot for these data and point out any clusters or outliers.
2.64 Reconsider Exercise 2.63. The following data give the number of times each of the 20 randomly
selected female students from the same state university ate at fast-food restaurants during the same 7-day
period.
0424 10 2505
6 114624560
a. Create a dotplot for these data.
b. Use the dotplots for male and female students to compare the two data sets.
2.65 The following table gives the number of times each of the listed players of the 2008 Philadelphia
Phillies baseball team was hit by a pitch (HBP). The list includes all players with at least 150 at-bats.
Player
HBP
Player
HBP
C. Utley
27
R. Howard
3
C. Coste
10
P. Burrell
1
S. Victorino
7
G. Dobbs
1
J. Rollins
5
G. Jenkins
1
C. Ruiz
4
P. Feliz
J. Werth
4
T. Iguchi
E. Bruntlett
3
Source: Major League
Baseball, 2008.
Create a dotplot for these data. Mention any clusters and/or outliers you observe.
USES AND MISUSES... TRUNCATING THE axes
Graphical analyses are an important part of statistics. However, graph-
ics software allows people to change the appearance of any graph.
Thus, one needs to be careful when reading a graph in order to in-
terpret the data properly. As an example, consider the following data
on the number of identity theft cases per 1000 people in eight states
during the year 2005.
State
Identity Theft Cases
per 1000 People
Arizona
156.9
Nevada
130.2
California
125.0
Texas
116.5
Colorado
97.2
Florida
95.8
Washington
92.4
New York
90.3
A proper bar graph of these data would look like the one shown in
Figure 2.21.
160
140
120
100
so
60
40
20
Chart of Identity Theft Cases vs State
Source: Identity Theft Security, 2006.
Figure 2.21 Identity theft cases for eight states.
62 Chapter 2 Organizing and Graphing Data
In addition to being able to obtain approximate rates by look-
ing at the vertical axis, we can determine the relative sizes of the
rates by looking at the bars. For example, we can see that the rates
in Colorado and Florida are approximately the same, whereas the
rate in Arizona is approximately 50% higher than the rates in Wash-
ington and New York.
We can present the same data using a bar graph similar to the
one shown in Figure 2.22. The data have not been altered in this
graph. However, the vertical axis has been changed so that it begins
at 80 instead of at zero. As a result, many people will look at this
graph and see very different relationships among the states than the
relationships that actually exist. For example, the bar representing
Arizona appears to be eight times the height of the bar for New York.
This would correspond to an identity theft rate that is 700% higher
for Arizona than for New York when it is actually only 50% higher.
Similarly, the rate for Texas appears to be 300% higher than for
New York when it is really only slightly more than 30% higher.
that the above data on identity thefts were displayed in a pie chart
such as the one shown here.
Figure 2.22 Identity theft cases for eight states.
There are many reasons why the second version of the bar chart
might be used. In some circumstances, the creator of the graph may
not know any better. In other cases, it could be that the user has
ulterior motives, such as to justify increasing funding for fighting iden-
tity theft or to make a state or a region to appear to have a lower
rate than what it truly is.
Problematic graphs also occur because the user does not under-
stand the purpose of a specific type of graph. For example, suppose
Pie Chart of Identity Theft Cases vs State
1 Floret*
MY&rfc
There are many possible misinterpretations of the data as a
result of using this pie chart. For example, a person might look at
the graph and believe that 17.4% of all identity theft cases occur
in Arizona. There are two major problems with this interpretation.
First, only eight states are represented in this chart, so all of the
percentages are overstated with respect to the entire United States.
Second, and more important, the percentages are based on the
sums of the state rates per 1000 people. This cannot be done un-
less all of the states have the same population.
California and Arizona can be used to illustrate the latter point.
If you look at the pie chart, it would appear that the number of
identity theft cases in Arizona is higher than the number of cases
in California. However, the rate of identity theft is affected by the
population size. California had 45,175 cases, while Arizona had
9320. The 2005 estimated populations of the two states were
36.1 million in California and 5.9 million in Arizona. California had
slightly less than five times as many cases as Arizona, but more than
six times as many people, which is why Arizona's rate is higher than
California's rate.
As you are learning about different types of graphs in statis-
tics, it is important to know how to make them and how to inter-
pret them; otherwise you can be misled or deceived by them.
Glossary
Bar graph A graph made of bars whose heights represent the fre-
quencies of respective categories.
Class An interval that includes all the values in a (quantitative)
data set that fall within two numbers, the lower and upper limits of
the class.
Class boundary The midpoint of the upper limit of one class and
the lower limit of the next class.
Class frequency The number of values in a data set that belong
to a certain class.
Class midpoint or mark The class midpoint or mark is obtained
by dividing the sum of the lower and upper limits (or boundaries)
of a class by 2.
Class width or size The difference between the two boundaries of
a class.
Supplementary Exercises 63
Cumulative frequency The frequency of a class that includes
all values in a data set that fall below the upper boundary of that
class.
Cumulative frequency distribution A table that lists the total
number of values that fall below the upper boundary of each class.
Cumulative percentage The cumulative relative frequency multi-
plied by 100.
Cumulative relative frequency The cumulative frequency of a
class divided by the total number of observations.
Frequency distribution A table that lists all the categories or
classes and the number of values that belong to each of these cate-
gories or classes.
Grouped data A data set presented in the form of a frequency
distribution.
Histogram A graph in which classes are marked on the horizon-
tal axis and frequencies, relative frequencies, or percentages are
marked on the vertical axis. The frequencies, relative frequencies, or
percentages of various classes are represented by bars that are drawn
adjacent to each other.
Ogive A curve drawn for a cumulative frequency distribution.
Outliers or Extreme values Values that are very small or very
large relative to the majority of the values in a data set.
Percentage The percentage for a class or category is obtained by
multiplying the relative frequency of that class or category by 100.
Pie chart A circle divided into portions that represent the relative fre-
quencies or percentages of different categories or classes.
Polygon A graph formed by joining the midpoints of the tops of
successive bars in a histogram by straight lines.
Raw data Data recorded in the sequence in which they are col-
lected and before they are processed.
Relative frequency The frequency of a class or category divided
by the sum of all frequencies.
Skewed-to-the-left histogram A histogram with a longer tail on
the left side.
Skewed-to-the-right histogram A histogram with a longer tail on
the right side.
Stem-and-leaf display A display of data in which each value is
divided into two portions — a stem and a leaf.
Symmetric histogram A histogram that is identical on both sides
of its central point.
Ungrouped data Data containing information on each member of
a sample or population individually.
Uniform or rectangular histogram A histogram with the same
frequency for all classes.
Supplementary Exercises
2.66 The following data give the political party of each of the first 30 U.S. presidents. In the data, D stands
for Democrat, DR for Democratic Republican, F for Federalist, R for Republican, and W for Whig.
F
F
DR
DR
DR
DR
D
D
W
W
D
W
W
D
D
R
D
R
R
R
R
D
R
D
R
R
R
D
R
R
a. Prepare a frequency distribution table for these data.
b. Calculate the relative frequency and percentage distributions.
c. Draw a bar graph for the relative frequency distribution and a pie chart for the percentage
distribution.
d. What percentage of these presidents were Whigs?
2.67 In a November 2008 Harris Poll, U.S. adults were asked "Will the Obama administration be too lib-
eral or conservative?" Of the respondents, 35% said that it will be too liberal (L), 43% said that it will be
neither too liberal nor too conservative (N), 4% said that it will be too conservative (C), and 18% said that
they do not know (K). In a recent poll, 40 people were asked whether the Obama administration has been
too liberal or too conservative. Their responses are given below.
L
N
K
K
C
L
K
K
L
L
K
K
L
K
N
N
N
K
N
K
N
K
K
N
L
L
N
N
K
K
L
K
L
N
L
L
N
N
K
K
a. Prepare a frequency distribution for these data.
b. Calculate the relative frequencies and percentages for all classes.
c. Draw a bar graph for the frequency distribution and a pie chart for the percentage distribution.
d. What percentage of these respondents said "too liberal"?
2.68 The following data give the numbers of television sets owned by 40 randomly selected households.
1 1 2 3 2 4 1 3 2 1
3021232322
12 1113 1112
2423131224
64 Chapter 2 Organizing and Graphing Data
a. Prepare a frequency distribution table for these data using single-valued classes.
b. Compute the relative frequency and percentage distributions.
c. Draw a bar graph for the frequency distribution.
d. What percentage of the households own two or more television sets?
2.69 Twenty-four students from universities in Connecticut were asked to name the five current members
of the U.S. House of Representatives from Connecticut. The number of correct names supplied by the stu-
dents are given below.
423554315443
532313252150
a. Prepare a frequency distribution for these data using single- valued classes.
b. Compute the relative frequency and percentage distributions.
c. What percentage of the students in this sample named fewer than two of the representatives
correctly?
d. Draw a bar graph for the relative frequency distribution.
2.70 The following data give the amounts spent on video rentals (in dollars) during 2009 by 30 house-
holds randomly selected from those who rented videos in 2009.
595
24
6
100
100
40
622
405
90
55
155
760
405
90
205
70
180
88
808
100
240
127
83
310
350
160
22
111
70
15
a. Construct a frequency distribution table. Take $1 as the lower limit of the first class and $200
as the width of each class.
b. Calculate the relative frequencies and percentages for all classes.
c. What percentage of the households in this sample spent more than $400 on video rentals
in 2009?
2.71 The following data give the numbers of orders received for a sample of 30 hours at the Timesaver
Mail Order Company.
34
44
31
52
41
47 38
35
32
39
28
24
46
41
49
53 57
33
27
37
30
27
45
38
34
46 36
30
47
50
a. Construct a frequency distribution table. Take 23 as the lower limit of the first class and 7 as
the width of each class.
b. Calculate the relative frequencies and percentages for all classes.
c. For what percentage of the hours in this sample was the number of orders more than 36?
2.72 The following data give the amounts (in dollars) spent on refreshments by 30 spectators randomly
selected from those who patronized the concession stands at a recent Major League Baseball game.
4.95
27.99
8.00
5.80
4.50
2.99
4.85
6.00
9.00
15.75
9.50
3.05
5.65
21.00
16.60
18.00
21.77
12.35
7.75
10.45
3.85
28.45
8.35
17.70
19.50
11.65
11.45
3.00
6.55
16.50
a. Construct a frequency distribution table using the less-than method to write classes. Take $0 as
the lower boundary of the first class and $6 as the width of each class.
b. Calculate the relative frequencies and percentages for all classes.
c. Draw a histogram for the frequency distribution.
2.73 The following data give the repair costs (in dollars) for 30 cars randomly selected from a list of cars
that were involved in collisions.
2300
750
2500
410
555
1576
2460
1795
2108
897
989
1866
2105
335
1344
1159
1236
1395
6108
4995
5891
2309
3950
3950
6655
4900
1320
2901
1925
6896
a. Construct a frequency distribution table. Take $1 as the lower limit of the first class and $1400
as the width of each class.
b. Compute the relative frequencies and percentages for all classes.
c. Draw a histogram and a polygon for the relative frequency distribution.
d. What are the class boundaries and the width of the fourth class?
Supplementary Exercises 65
2.74 Refer to Exercise 2.70. Prepare the cumulative frequency, cumulative relative frequency, and
cumulative percentage distributions by using the frequency distribution table of that exercise.
2.75 Refer to Exercise 2.71. Prepare the cumulative frequency, cumulative relative frequency, and cumulative
percentage distributions using the frequency distribution table constructed for the data of that exercise.
2.76 Refer to Exercise 2.72. Prepare the cumulative frequency, cumulative relative frequency, and
cumulative percentage distributions using the frequency distribution table constructed for the data of
that exercise.
2.77 Construct the cumulative frequency, cumulative relative frequency, and cumulative percentage dis-
tributions by using the frequency distribution table constructed for the data of Exercise 2.73.
2.78 Refer to Exercise 2.70. Prepare a stem-and-leaf display for the data of that exercise.
2.79 Construct a stem-and-leaf display for the data given in Exercise 2.71.
2.80 The following table gives the 2008 endowments (in billions of dollars) for the six U.S. universities
that had the largest endowments.
Endowment
University (billions of dollars)
Harvard University 36.6
Yale University 22.9
Stanford University 17.2
Princeton University 16.3
University of Texas System 16.1
Massachusetts Institute of Technology 10.1
Source: National Association of College and University Business Officers, 2008.
Draw two bar graphs for these data — the first without truncating the axis on which endowments are marked,
and the second by truncating this axis. In the second graph, mark the endowments on the vertical axis
starting with $9 billion. Briefly comment on the two bar graphs.
2.81 The following table lists the average price per gallon for unleaded regular gasoline in the United
States from 1999 to 2008.
Year
Average Price per Gallon
(dollars)
1999
1.136
2000
1.484
2001
1.420
2002
1.345
2003
1.561
2004
1.852
2005
2.270
2006
2.572
2007
2.796
2008
3.246
Source: Energy Information Administration, 2008.
Draw two bar graphs for these data — the first without truncating the axis on which price is marked, and
the second by truncating this axis. In the second graph, mark the prices on the vertical axis starting with
$1.00. Briefly comment on the two bar graphs.
2.82 Reconsider the data on the times (in minutes) taken to commute from home to work for 20 workers
given in Exercise 2.53. Create a dotplot for those data.
2.83 Reconsider the data on the numbers of orders received for a sample of 30 hours at the Timesaver
Mail Order Company given in Exercise 2.71. Create a dotplot for those data.
66 Chapter 2 Organizing and Graphing Data
2.84 Twenty-four students from a university in Oregon were asked to name the five current members of
the U.S. House of Representatives from their state. The following data give the numbers of correct names
given by these students.
551245315501
235431525453
Create a dotplot for these data.
2.85 The following data give the numbers of visitors during visiting hours on a given evening for each of
the 20 randomly selected patients at a hospital.
3 1 4 2 4 1 1 3
4202221 130
Create a dotplot for these data.
Advanced Exercises
2.86 The following frequency distribution table gives the age distribution of drivers who were at fault in
auto accidents that occurred during a 1-week period in a city.
Age (years) /
18 to less than 20 7
20 to less than 25 12
25 to less than 30 18
30 to less than 40 14
40 to less than 50 15
50 to less than 60 16
60 and over 35
a. Draw a relative frequency histogram for this table.
b. In what way(s) is this histogram misleading?
c. How can you change the frequency distribution so that the resulting histogram gives a clearer
picture?
2.87 Refer to the data presented in Exercise 2.86. Note that there were 50% more accidents in the 25 to
less than 30 age group than in the 20 to less than 25 age group. Does this suggest that the older group of
drivers in this city is more accident-prone than the younger group? What other explanation might account
for the difference in accident rates?
2.88 Suppose a data set contains the ages of 135 autoworkers ranging from 20 to 53 years.
a. Using Sturge's formula given in footnote 1 on page 37, find an appropriate number of classes
for a frequency distribution for this data set.
b. Find an appropriate class width based on the number of classes in part a.
2.89 Statisticians often need to know the shape of a population to make inferences. Suppose that you are
asked to specify the shape of the population of weights of all college students.
a. Sketch a graph of what you think the weights of all college students would look like.
b. The following data give the weights (in pounds) of a random sample of 44 college students
(F and M indicate female and male, respectively).
123 F
195 M
138 M
115 F
179 M
119 F
148 F
147 F
180 M
146 F
179 M
189 M
175 M
108 F
193 M
114 F
179 M
147 M
108 F
128 F
164 F
174 M
128 F
159 M
193 M
204 M
125 F
133 F
115 F
168 M
123 F
183 M
116 F
182 M
174 M
102 F
123 F
99 F
161 M
162 M
155 F
202 M
110 F
132 M
i. Construct a stem-and-leaf display for these data.
ii. Can you explain why these data appear the way they do?
c. Now sketch a new graph of what you think the weights of all college students look like. Is this
similar to your sketch in part a?
Supplementary Exercises 67
2.90 Consider the two histograms given in Figure 2.23, which are drawn for the same data set. In this
data set, none of the values are integers.
Histogram of CI
10
u
E
Histogram of CI
4 6
CI
10
Figure 2.23 Two histograms for the same data.
a. What are the endpoints and widths of classes in each of the two histograms?
b. In the first histogram, of the observations that fall in the interval that is centered at 8, how
many are actually between the left endpoint of that interval and 8? Note that you have to
consider both histograms to answer this question.
c. Observe the leftmost bars in both histograms. Why is the leftmost bar in the first histogram
misleading?
2.91 Refer to the data on weights of 44 college students given in Exercise 2.89. Create a dotplot of all 44
weights. Then create stacked dotplots for the weights of male and female students. Describe the similar-
ities and differences in the distributions of weights of male and female students. Using all three dotplots,
explain why you cannot distinguish the lightest males from the heaviest females when you consider only
the dotplot of all 44 weights.
2.92 The pie chart in Figure 2.24 shows the percentage distribution of ages (i.e., the percentages of
all prostate cancer patients falling in various age groups) for men who were recently diagnosed with
prostate cancer.
a. Are more or fewer than 50% of these patients in their 50s? How can you tell?
b. Are more or fewer than 75% of these patients in their 50s and 60s? How can you tell?
68 Chapter 2 Organizing and Graphing Data
Pie Chart of age group
Figure 2.24 Pie chart of age groups.
c. A reporter looks at this pie chart and says, "Look at all these 50-year-old men who are getting
prostate cancer. This is a major concern for a man once he turns 50." Explain why the reporter
cannot necessarily conclude from this pie chart that there are a lot of 50-year-old men with
prostate cancer. Can you think of any other way to present these cancer cases (both graph and
variable) to determine if the reporter's claim is valid?
2.93 Stem-and-leaf displays can be used to compare distributions for two groups using a back-to-back
stem-and-leaf display. In such a display, one group is shown on the left side of the stems, and the other
group is shown on the right side. When the leaves are ordered, the leaves increase as one moves away
from the stems. The following stem-and-leaf display shows the money earned per tournament entered for
the top 30 money winners in the 2008-09 Professional Bowlers Association men's tour and for the top 21
money winners in the 2008-09 Professional Bowlers Association women's tour.
Women's
Men's
8
8871
1
65544330
2
334456899
840
3
03344678
52
4
011237888
21
5
9
6
9
5
7
8
7
9
5
The leaf unit for this display is 100. In other words, the data used represent the earnings in hundreds of
dollars. For example, for women's tour, the first number is 08, which is actually 800. The second number
is 11, which actually is 1100.
a. Do the top money winners, as a group, on one tour (men's or women's) tend to make more
money per tournament played than on the other tour? Explain how you can come to this
conclusion using the stem-and-leaf display.
b. What would be a typical earnings level amount per tournament played for each of the two tours?
c. Do the data appear to have similar spreads for the two tours? Explain how you can come to this
conclusion using the stem-and-leaf display.
d. Does either of the tours appears to have any outliers? If so, what are the earnings levels for
these players?
Supplementary Exercises 69
2.94 The following table lists the earnings per event that were referred to in Exercise 2.93. Although the
table lists earnings per event, players are listed in order of their total earnings, not their earnings per event.
Note that men and women are ranked together in the table.
Name
Earnings per Event
(in dollars)
Name
Earnings per Event
(in dollars)
Txlr\i"m Tlnlrp
7JUO.U /
p lifTPtiP Mi^i^nnp
JH/UgCllC 1V1UV_,U.11C
94.75 88
Wpc A/Talntt
wt^a ±\±aL\Jii
879S 61
Rnnnip Rii*;qp11
2540 63
Pot n r*V A 1 1 pst-i
6Q7Q 41
1? i tr*tii p A 1 lpn
??4n nn
irifTC R artiPQ
S970 00
TapV Tnrpk"
J lILK ,1 Lll L IS.
2322.94
W/nlfpv Pqu \A/illinmc It*
vv ciiLCi JVdy vv liiidiiiA j i.
47S8 87
T 17 \r\rt ti cr^Ti
7snn nn
Rill O'Neill
4884 18
IVTipIipIIp PplHman
IvlldlL 11L 1 11.11 1 ltll 1
5214.29
T?nitir\ P'l fr p
iviiliiu rage
487? SO
i ni"r\1\/n 1 ~1 Mnn_ R nil qtH
v^aHJlyll -L/tJllU JJOllctlU
51 85 71
John Nolen
4801.56
Stclanie Nation
4542.86
IVllJvC OL1 UgglllS
tju / .UU
Tpfl T1 1 TPl" PptVIpL"
J C111111CI .F CL1 ll<Iv
4285.71
Rrnrl An<rp1r\
JJld.Ll rtllgCHJ
429 1.18
Tnni AA/npccnpi"
J UU1 VV UCoallCl
1885 71
Rpfp W/p* r^^T"
r CLC VV C UCL
4HS 90
v n cinTir\n Pin nfm/clri/
Parlrpr Rnhn TTT
I til KCI JDUIlll 111
4101.47
lV^iccv Rpllinrlpi"
iviisay .dciiiiiu.ci
7^86 75
IVllL-lldCl ragall
^851 76
Tiifinrli"ci A cnihr
IvldllUl a rt.SUd.ly
954? 86
S tp vp T-f n nn an
JIL VL llUllllilll
4035 63
Tncha RpiH
111, Mill IvtlU
?4nn nn
Z.'-rVJVJ . \J\J
TnmiTu; Tnnpc
1U111111 y J UUCft
J / 1J.OO
\A/pnrl\7 \\A Qcnnprenr
VVCllUy lVld.UUllCl ftUll
105fS 00
Flatinv WKPmari
L'ciiiiiy vv i >*w iiiuii
f^lar:! friiprrprn
?4<S(S 67
3399.41
Shalin 7nlkifli
Olltlllll ^jUIIVIIII
2098.57
rVTik"fi TCmviiniPTTii
IVXIJVtl 1VU1 V LI 111^1111
3396.47
Tpnnpllp lVTillipati
1 I. 1 1 1 1 k. 1 1 k. 1V111I1_,U1
2331.67
Jeff Carter
3410.94
Shannon O'Keefe
2640.00
Michael Machuga
3455.33
Joy Esterson
1807.14
Ryan Shafer
2983.53
Adrienne Miller
1798.57
Mike Wolfe
2902.35
Brenda Mack
1833.33
Steve Jaros
2884.12
Olivia Sandham
1100.00
Chris Loschetter
3035.63
Amy Stolz
2500.00
Mike DeVaney
2681.76
Kelly Kulick
830.00
Ken Simard
2412.19
Source: Professional Bowlers Associaton, April 13, 2009.
A graph that is similar to an ogive is a graph of the empirical cumulative distribution function (CDF). The
primary difference between an ogive and an empirical CDF is that the empirical CDF looks like a set of
steps, as opposed to a set of slanted lines. The height of each step corresponds to the percentage of
observations that occur at a specific value. Longer (not higher) steps occur when there are bigger gaps
between observations.
a. Figures 2.25(a) and (b) contain the empirical CDFs of the earnings per event for the two tours
(men's and women's), in some order. In other words, one of these two figures is for the men's
tour, and the other is for the women's tour, but not in that order necessarily. Match the CDFs to
the respective tours. Give three reasons for your choices.
b. Both distributions are skewed to the right. Use the information about longer steps to explain
why the distributions are skewed to the right.
c. What are the approximate values of the CDFs corresponding to $3000 per tournament played and
$4000 per tournament played? Based on this information, what is the approximate percentage of
bowlers who earned between $3000 and $4000 per tournament played?
2.95 Table 2.18 contains the differences in the obesity rates (called rate change in the table) for the years be-
tween 2007 and 1997 for each of the 50 states and the District of Columbia. The obesity rate is the percent-
age of people having a body mass index (BMI) of 30 or higher. Figure 2.26 contains a dotplot of these data,
a. Analyze the dotplot carefully. What value would you provide if asked to report a "typical" obe-
sity rate change? Why did you choose this value?
70 Chapter 2 Organizing and Graphing Data
Figure 2.25(a)
Figure 2.25(6)
Dot plot of Obesity Rate Change
S Ifaaaaflaaal • •
— * • mjf a a#a*aaa*a*aaa* » • ■ ■ *L_
6.0 7.2 8.4 9.6 10.8 12.0 13.2
Obesity Rate Change
Figure 2.26 Dotplot of obesity rate changes (year 2007 minus year 1997).
Supplementary Exercises 71
Table 2.18 Difference in 2007 and 1997 Obesity Rates, by State
Rate
Rate
Rate
State
Change
State
Change
State
Change
AL
12.1
KY
5.6
ND
9.5
AK
7.8
LA
in t
10.2
OH
9.8
AZ
13
ME
8.6
OK
13
AK
11).
MD
/.y
UK
6. 1
CA
6.6
MA
6.5
FA
9.6
CO
6.9
MI
O 1
8.4
Kl
7.6
C 1
6.5
MN
9. 1
SC
i 1 c
11.5
Dh
5.6
A /TC
Ma
1 A
cn
SU
9.2
D.C.
7..i
MO
8.4
TN
IZA
FL
7.5
MT
7.2
TX
9.4
A
1 3.0
MX?
9
U 1
6.6
nl
7 S
NV
1
VT
V 1
ID
8.2
NH
10.2
VA
7.9
IL
7.8
NJ
7.5
WA
10.1
IN
5.6
NM
9.1
WV
8.9
IA
7.5
NY
9
WI
8.1
KS
12.2
NC
9.7
WY
8.7
Source: Centers for Disease Control and Prevention, July 24, 2008.
b. What number do you feel most accurately represents the number of outliers in this data set: 0,
1, 3, 4, 6, 9, or 10? Explain your reasoning, including the identification of the observations, if
any, that you feel are outliers.
c. Would you classify this distribution as being skewed to the left, skewed to the right, or approxi-
mately symmetric? Explain.
d. The largest increase in the obesity rate during this period took place in Georgia (13.8), whereas
the smallest increase took place in Vermont (5.4). Explain why this information should not lead
you to conclude that Georgia had the highest obesity rate in 2007 and that Vermont had the
lowest obesity rate in 2007. (Note: The highest and lowest obesity rates in 2007 were in
Mississippi and Colorado, respectively.)
2.96 Figure 2.27 contains stacked dotplots of 2007 state obesity rates by different geographic regions —
Midwest, Northeast, South, and West.
Dotplot of 2007 Obesity Rate
Midwest
Northeast
South
West
_l ■ — —
19.8
21.6
23.4 25.2 27.0
2007 Obesity Rate
28.8
30.6
Figure 2.27 2007 state obesity rates by geographic region.
72 Chapter 2 Organizing and Graphing Data
a. Which region has the least variability (greatest consistency) of obesity rates? Which region has
the most variability (least consistency) of obesity rates? Justify your choices.
b. Which region tends to have the highest obesity rates? Which region tends to have the lowest
obesity rates? Justify your choices.
c. Are there any regions that have at least one obesity rate that could be considered an outlier? If
so, specify the region(s) and the observation(s).
2.97 CBS Sports had a Facebook page for the 2009 NCAA Men's Basketball Tournament including bracket
contests, discussion sites, and a variety of polls. One of the polls asked users to identify their most de-
spised teams. The following pie chart (Figure 2.28) gives a breakdown of the votes by the conference of
the most despised teams as of 10:53 EDT on March 16, 2009.
Pie Chart of Conference
Big Ten
Figure 2.28 Pie chart of conference of the most despised NCAA men's
basketball team.
a. Are there any conferences that received more than 25% of the votes? If so, which conference(s)?
How can you tell?
b. Which two conferences appear to have the closest percentages of the votes?
c. A bar chart for the same data is presented in Figure 2.29. Comparing the bar chart to the pie
chart, match the conferences to the bars. In other words, explain which bar represents which
conference.
Chart of Votes
90000
80000-
70000"
60000
s
■g 5O0M
4O000 i 1
30000
2MO0
Tn I.M.I I
abcdefgh
Conference
Figure 2.29 Bar chart of conference of the most despised NCAA
men's basketball team.
Self-Review Test 73
Self-Review Test
1. Briefly explain the difference between ungrouped and grouped data and give one example of each
type.
2. The following table gives the frequency distribution of times (to the nearest hour) that 90 fans spent
waiting in line to buy tickets to a rock concert.
Waiting Time
(hours)
Frequency
Oto 6
5
7 to 13
27
14 to 20
30
21 to 27
20
28 to 34
8
Circle the correct answer in each of the following statements, which are based on this table.
a. The number of classes in the table is 5, 30, 90.
b. The class width is 6, 7, 34.
c. The midpoint of the third class is 16.5, 17, 17.5.
d. The lower boundary of the second class is 6.5, 7, 7.5.
e. The upper limit of the second class is 12.5, 13, 13.5.
f. The sample size is 5, 90, 11.
g. The relative frequency of the second class is .22, .41, .30.
3. Briefly explain and illustrate with the help of graphs a symmetric histogram, a histogram skewed to
the right, and a histogram skewed to the left.
4. Twenty elementary school children were asked if they live with both parents (B), father only (F),
mother only (M), or someone else (S). The responses of the children follow.
MBBMF SBMFM
B FBMMBBF BM
a. Construct a frequency distribution table.
b. Write the relative frequencies and percentages for all categories.
c. What percentage of the children in this sample live with their mothers only?
d. Draw a bar graph for the frequency distribution and a pie chart for the percentages.
5. A large Midwestern city has been chronically plagued by false fire alarms. The following data set
gives the number of false alarms set off each week for a 24-week period in this city.
10 4 8 7 3 7 10 2 6 12 11 8
1 6 5 13 9 7 5 1 14 5 15 3
a. Construct a frequency distribution table. Take 1 as the lower limit of the first class and 3 as the
width of each class.
b. Calculate the relative frequencies and percentages for all classes.
c. What percentage of these weeks had 9 or fewer false alarms?
d. Draw the frequency histogram and polygon.
6. Refer to the frequency distribution prepared in Problem 5. Prepare the cumulative percentage distri-
bution using that table. Draw an ogive for the cumulative percentage distribution.
7. Construct a stem-and-leaf display for the following data, which give the times (in minutes) 24 cus-
tomers spent waiting to speak to a customer service representative when they called about problems
with their Internet service provider.
12 15 7 29 32 16 10 14 17 8 19 21
4 14 22 25 18 6 22 16 13 16 12 20
74 Chapter 2 Organizing and Graphing Data
8. Consider this stem-and-leaf display:
3
3
7
4
2 4
6
7
5
1 3
3
6
6
7
7
7
1 9
Write the data set that was used to construct this display.
9. Make a dotplot for the data given in Problem 5.
Mini-Projects
■ MINI-PROJECT 2-1
Using the data you gathered for the mini-project in Chapter 1, prepare a summary of that data set that in-
cludes the following.
a. Prepare an appropriate type of frequency distribution table for one of the quantitative variables
and then compute relative frequencies and cumulative relative frequencies.
b. Create a histogram, a stem-and-leaf display, and a dotplot of the data. Comment on any symme-
try or skewness and on the presence of clusters and any potential outliers.
c. Make stacked dotplots of the same variable (as in parts a and b) based on the values of one of
your categorical variables. For example, if your quantitative variable is GPAs of students, your cat-
egorical variable could be gender. Comment on the similarities and differences between the dis-
tributions for the different values of your categorical variable.
■ MINI-PROJECT 2-2
Choose 15 of each of two types of magazines (news, sports, fitness, entertainment, and so on) and record
the percentage of pages that contain at least one advertisement. Using these percentages and the types of
magazines, write a brief report that covers the following:
a. Prepare an appropriate type of frequency distribution table for the quantitative variable and then
compute relative frequencies and cumulative relative frequencies.
b. Create a histogram, a stem-and-leaf plot, and a dotplot of all of the data. Comment on any sym-
metry or skewness, as well as the presence of clusters and any potential outliers.
c. Make stacked dotplots of the same variable for each of the two types of magazines. Comment on
the similarities and differences between the distributions for the two types of magazines.
DECIDE FOR YOURSELF
Deciding About Statistical Properties
Look around you. Graphs are everywhere. Business reports, news-
papers, magazines, and so forth are all loaded with graphs.
Unfortunately, some people feel that the primary purpose of
graphs is to provide a break from the humdrum text. Executive sum-
maries will often contain graphs so that CEOs and executive vice
presidents need only to glance at these graphs to assume that they
understand everything without reading more than a paragraph or so of
the report. In reality, the usefulness of graphs is somewhere between
the fluff of the popular press and the quick answer of the boardroom.
Here you are asked to interpret some graphs, primarily by using
them to compare distributions of a variable. As we will discuss in
Chapter 3, some of our concerns have to do with the location of the
center of a distribution and the variability or spread of a distribution.
We can use graphs to compare the centers and variability of two or
more distributions.
In practice, the graphs are made using statistical software, so it
is important to recognize that computer software is programmed to
use the same format for each graph of a specific type, unless you tell
the software to do differently. For example, consider the two his-
tograms in Figures 2.30 and 2.31 that are drawn for two different
data sets.
1. Examine the two graphs of Figures 2.30 and 2.31.
2. Explain what is meant by the statement "the shapes of the two
distributions are the same."
3. Does the fact that the shapes of the two distributions are the
same imply that the centers of the two distributions are the same?
Why or why not? Explain.
4. Does the fact that the shapes of the two distributions are the
same imply that the spreads of the two distributions are the same?
Why or why not? Explain.
Technology Instruction 75
Figure 2.30 Histogram of data temp 1.
5. It turns out that the same variable was represented in the two
graphs but with different units of measurement. Can you figure out
the units?
Another situation that is important to compare is when two
graphs cover a similar range but have different shapes, such as the
histograms in Figures 2.32 and 2.33.
Histogram of Example 2a
Example 2a
Figure 2.32 Histogram of example 2a.
Figure 2.31 Histogram of data temp 2.
1. Examine the two histograms of Figures 2.32 and 2.33.
2. These two distributions have the same center but do not have
the same spread. Decide which distribution has the larger spread and
explain the reasoning behind your decision.
Answer all the above questions again after reading Chapter 3.
Histogram of Example 2b
Example 2b
Figure 2.33 Histogram of example 2b.
4f
ECHNOLOGY
INSTRUCTION
jmS fVAZ PlOtS
aSroff
Type: L^: Li
>o<- >njH
Xlist:|_i
Fre-=i: 1_
Screen 2.1
Organizing Data
1. To create a frequency histogram for a list of data, press STAT PLOT, which you access
by pressing 2nd > Y=. The Y= key is located at the top left of the calculator buttons.
2. Make sure that only one plot is turned on. If more than one plot is turned on, you can turn
off the unwanted plots by using the following steps. Press the number corresponding to the
plot you wish to turn off. A screen similar to Screen 2.1 will appear. Use the arrow keys
to move the cursor to the Off button, then press ENTER. Now use the arrow keys to move
76_ Chapter 2 Organizing and Graphing Data
to the row with Plotl, Plot2, and Plot3. If there is another plot that you need to turn off,
select that plot by moving the cursor to that plot, pressing ENTER, and repeating the previ-
ous procedure. If not, move the cursor to the plot you wish to use and press ENTER.
3. In the Type rows, use the right arrow to move to the third column in the first row that
looks like a histogram, and press Enter. Move to Xlist to enter the name of the list where
the data are located. Press 2nd > Stat, then use the up and down arrows to move through
the list names until you find the list you want to use. Press ENTER. Leave the Freq set-
ting at 1. (Note: if you are using one of the lists named LI, L2, L3, L4, L5, or L6, you
can enter the list name by pressing 2nd followed by one of the numbers 1 through 6, as
they correspond to the list names LI through L6.)
4. To see the graph, select ZOOM > 9 (the ZOOMSTAT function), where ZOOM is the
third key in the top row. This sets the window settings to display your graph.
5. If you would like to change the class width and/or the starting point of the first inter-
val, select WINDOW (see Screen 2.2). To change the class width, change the value of
Xscl to the desired width. To change the starting point of the first interval, change the
value of Xmin to the desired point. Press GRAPH, which is the fifth button in the top
row. (Note: After making either or both of these changes, you may need to change the
values of Xmax and Ymax to see the entire graph. The difference between Xmax and
Xmin should be a multiple of Xscl. As an example, if Xmin = 5 and Xscl = 10 and
the largest data point is 93, then Xmax should be set to 95 because 95 — 5 = 90,
which is a multiple of 10, and 95 is larger than the largest data point. The purpose of
changing Ymax is to be able to see the tops of the bars of the histogram. If the bars
run off the top of the calculator screen, increase Ymax, and press GRAPH.)
6. If you would like to see the interval endpoints and the number of observations in each class
(which is given by the height of the corresponding bar), press TRACE, then use the left
and right arrows to move from one bar to the next. When you are done, press CLEAR.
The functions for creating many common graphs are listed in the pulldown menu Graph.
The following instructions will demonstrate how to use Minitab to create two types of graphs
for categorical variables — a bar chart and a pie chart — and three types of graphs for quantita-
tive variables — a frequency histogram, a stem-and-leaf display, and a dotplot.
Bar Chart
1. If you have raw categorical data entered in a column (such as CI), select Graph > Bar
Chart. In the resulting dialog box, select Bars Represent: Counts of unique values and
Simple. Click OK. In the new dialog box, type CI in the box below Categorical Variables
and click OK.
2. If you have categorical data in a frequency table, with the categories entered in CI and the
frequencies in C2, select Graph > Bar Chart. In the resulting dialog box, select Bars
Represent: Values from a table and Simple. Click OK. In the new dialog box, type C2
in the box below Graph variables and CI in the box below Categorical Variable and
click OK.
Pie Chart
1. If you have raw categorical data entered in CI, select Graph > Pie Chart. In the resulting
dialog box, select Chart raw data, type CI in the box below Categorical Variables, and
click OK.
WINDOW
Xmin=5_
Xnax=95
Xscl=19
Vnin- -2. 16483
Vnax=15
Yscl=l
Xres=l
Screen 2.2
Technology Instruction 77
2. If you have categorical data in a frequency table, with the categories entered in CI and the
frequencies in C2, select Graph > Pie Chart. In the resulting dialog box, select Chart
values from a table, type C2 in the box below Summary variables and CI in the box
below Categorical Variable, and click OK.
Histogram - Simple
ci
I
Graph variables:
CI
Scale.,.
Labels,,
Data View,,
Multiple Graphs.,
Data Options,,
Help
OK
Cancel
Frequency Histogram
For a quantitative data set entered in CI, select
Graph > Histogram, select Simple, and click
OK. In the resulting dialog box, type CI in the
box below Graph Variables (see Screen 2.3) and
click OK. Minitab will produce a separate window
that contains the histogram (see Screen 2.4).
Screen 2.3
Screen 2.4
QUI
Stem -and -Leaf Display: C I
Stem- and- leaf of CI U = 18
Leaf Unit = 1.0
3 122
9 556689
9 1 133
6 1 557S9
12 4
< Nl| j
Screen 2.5
Stem-and-Leaf Display
For a quantitative data set entered in CI, select Graph > Stem-and-Leaf, type CI in
the box below Graph Variables, and click OK. The display will appear in the Session
window (see Screen 2.5).
Dotplot
For a quantitative data set entered in CI, select Graph > Dotplot, select the appro-
priate dotplot from the choices, and click OK. In the resulting dialog box, type CI
in the box below Graph Variables and click OK. The dotplot will appear in a new
window.
78 Chapter 2 Organizing and Graphing Data
m a I b
c
P
i
Data Boundaries
Frequencies
2
3
1.1
=FREQUENCY(a3:a20,b3
:b3}
4
2.1
5
_5_
?:/
10
6
5.2
is
7
US
20
8
6.3
25
9
6.6
1C
S.4
11
9.3
12
11
ia
14
13.4
15
15,1
16
15.7
17
17
IS
IS. 2
13
19.3
20
24
To create a frequency distribution for a range of numerical data in
Excel, decide how many categories you will have. Choose class bound-
aries between the categories so that you have one fewer boundary than
classes. Type the class boundaries into Excel.
Select where you want the class frequencies to appear, and select a
range of one more cell than the number of boundaries you have.
Type =frequency(.
Select the range of cells of numerical data, and then type a comma.
5. Select the range of class boundaries, and type a right parenthesis (see
Screen 2.6).
Screen 2.6
TECHNOLOGY ASSIGNMENTS
TA2.1 Construct a bar graph and a pie chart for the frequency distribution prepared in Exercise 2.5.
TA2.2 Construct a bar graph and a pie chart for the frequency distribution prepared in Exercise 2.6.
TA2.3 Refer to Data Set V that accompanies this text (see Preface and Appendix B) on the times taken
to run the Manchester Road Race for a sample of 500 participants. From that data set, select the 6th value,
and then select every 10th value after that (i.e., select the 6th, 16th, 26th, 36th, . . . values). This subsam-
ple will give you 50 measurements. (Such a sample selected from a population is called a systematic ran-
dom sample!) Construct a histogram for these data. Let the software you use decide on classes and class
limits.
TA2.4 Refer to Data Set I that accompanies this text on the prices of various products in different cities
across the country. Select a subsample of 60 from the column that contains information on pizza prices
and then construct a histogram for these data.
TA2.5 Construct a histogram for the data from Exercise 2.20 on the numbers of computer keyboards
assembled. Use the classes given in that exercise. Use the midpoints to mark the horizontal axis in the
histogram.
TA2.6 Prepare a stem-and-leaf display for the data given in Exercise 2.48.
TA2.7 Prepare a stem-and-leaf display for the data of Exercise 2.53.
TA2.8 Prepare a bar graph for the frequency distribution obtained in Exercise 2.28.
TA2.9 Prepare a bar graph for the frequency distribution obtained in Exercise 2.29.
TA2.10 Make a pie chart for the frequency distribution obtained in Exercise 2.19.
TA2.11 Make a pie chart for the frequency distribution obtained in Exercise 2.29.
TA2.12 Make a dotplot for the data of Exercise 2.64.
TA2.13 Make a dotplot for the data of Exercise 2.65.
During the 2008 season, among all baseball teams, the New York Yankees drew the highest av-
erage of spectators to the games. (See Case Study 3-1). Despite baseball being "America's National
Pastime," attendance at Major League Baseball games varies from team to team. What may attract
fans to baseball fields? Is it the number of championships won, market size, fan loyalty, or just love
for the team or game?
In Chapter 2 we discussed how to summarize data using different methods and to display data using
graphs. Graphs are one important component of statistics; however, it is also important to numerically
describe the main characteristics of a data set. The numerical summary measures, such as the ones
that identify the center and spread of a distribution, identify many important features of a distribution.
For example, the techniques learned in Chapter 2 can help us graph data on family incomes. However,
if we want to know the income of a "typical" family (given by the center of the distribution), the spread
of the distribution of incomes, or the relative position of a family with a particular income, the numeri-
cal summary measures can provide more detailed information (see Figure 3.1). The measures that
we discuss in this chapter include measures of (1) central tendency, (2) dispersion (or spread), and
(3) position.
3.1 Measures of Central
Tendency for Ungrouped
Data
Case Study 3-1 Average
Attendance at Baseball
Games
Case Study 3-2 The Gender
Pay Gap
3.2 Measures of Dispersion
for Ungrouped Data
3.3 Mean, Variance, and
Standard Deviation for
Grouped Data
3.4 Use of Standard
Deviation
Case Study 3-3 Here Comes
the SD
3.5 Measures of Position
3.6 Box-and-Whisker Plot
79
80 Chapter 3 Numerical Descriptive Measures
Figure 3.1
•* Spread
Center $76,260 Income
— Position of a
particular family
3.1 Measures of Central Tendency
for Ungrouped Data
We often represent a data set by numerical summary measures, usually called the typical val-
ues. A measure of central tendency gives the center of a histogram or a frequency distribu-
tion curve. This section discusses three different measures of central tendency: the mean, the
median, and the mode; however, a few other measures of central tendency, such as the trimmed
mean, the weighted mean, and the geometric mean, are explained in exercises following this
section. We will learn how to calculate each of these measures for ungrouped data. Recall from
Chapter 2 that the data that give information on each member of the population or sample
individually are called ungrouped data, whereas grouped data are presented in the form of a
frequency distribution table.
3.1.1 Mean
The mean, also called the arithmetic mean, is the most frequently used measure of central
tendency. This book will use the words mean and average synonymously. For ungrouped
data, the mean is obtained by dividing the sum of all values by the number of values in the
data set:
Sum of all values
Mean =
Number of values
The mean calculated for sample data is denoted by x (read as "x bar"), and the mean cal-
culated for population data is denoted by p (Greek letter mu). We know from the discussion
in Chapter 2 that the number of values in a data set is denoted by n for a sample and by N
for a population. In Chapter 1, we learned that a variable is denoted by x, and the sum of all
values of x is denoted by Xx Using these notations, we can write the following formulas for
the mean.
Calculating Mean for Ungrouped Data The mean for ungrouped data is obtained by
dividing the sum of all values by the number of values in the data set. Thus,
Xx
Mean for population data: p =
Mean for sample data: x = —
n
where Xx is the sum of all values, N is the population size, n is the sample size, p is the
population mean, and x is the sample mean.
3.1 Measures of Central Tendency for Ungrouped Data 81
■ EXAMPLE 3-1
Table 3.1 lists the total sales (rounded to billions of dollars) of six U.S. companies for 2008.
Table 3.1 2008 Sales of Six U.S. Companies
Calculating the sample mean
for ungrouped data.
Total Sales
Company (billions of dollars)
General Motors 149
Wal-Mart Stores 406
General Electric 183
Citigroup 107
Exxon Mobil 426
Verizon Communication 97
Find the 2008 mean sales for these six companies.
Solution The variable in this example is the 2008 total sales for a company. Let us denote
this variable by x. Then, the six values of x are
x x = 149, x 2 = 406, x 3 = 183, x 4 = 107, x 5 = 426, and x 6 = 97
where x x = 149 represents the 2008 total sales of General Motors, x 2 = 406 represents the
2008 total sales of Wal-Mart Stores, and so on. The sum of the 2008 sales for these six
companies is
Xx = X; + x 2 + x 3 + x 4 + x 5 + x 6
= 149 + 406 + 183 + 107 + 426 + 97 = 1368
Note that the given data include only six companies. Hence, they represent a sample. Because
the given data set contains six companies, n = 6. Substituting the values of Xx and n in the
sample formula, we obtain the mean 2008 sales of the six companies:
Xx 1368
x = — = = 228 = $228 Billion
n 6
Thus, the mean 2008 sales of these six companies was 228, or $228 billion. H
■ EXAMPLE 3-2
The following are the ages (in years) of all eight employees of a small company:
53 32 61 27 39 44 49 57
Find the mean age of these employees.
Solution Because the given data set includes all eight employees of the company, it repre-
sents the population. Hence, N = %. We have
Xx = 53 + 32 + 61 + 27 + 39 + 44 + 49 + 57 = 362
The population mean is
Xx 362
N 8
45.25 years
Thus, the mean age of all eight employees of this company is 45.25 years, or 45 years and
3 months.
Calculating the population
mean for ungrouped data.
82 Chapter 3 Numerical Descriptive Measures
Reconsider Example 3-2. If we take a sample of three employees from this company and
calculate the mean age of those three employees, this mean will be denoted by x. Suppose the
three values included in the sample are 32, 39, and 57. Then, the mean age for this sample is
32 + 39 + 57
= 42.67 years
x =
3
If we take a second sample of three employees of this company, the value of x will (most likely)
be different. Suppose the second sample includes the values 53, 27, and 44. Then, the mean age
for this sample is
53 + 27 + 44
x = = 41.33 years
Consequently, we can state that the value of the population mean ijl is constant. However, the
value of the sample mean x varies from sample to sample. The value of x for a particular sample
depends on what values of the population are included in that sample.
Sometime a data set may contain a few very small or a few very large values. As mentioned
in Chapter 2 on page 58, such values are called outliers or extreme values.
A major shortcoming of the mean as a measure of central tendency is that it is very sensitive
to outliers. Example 3-3 illustrates this point.
■ EXAMPLE 3-3
Table 3.2 lists the total philanthropic givings (in million dollars) by six companies during 2007.
Illustrating the effect of an
outlier on the mean.
Table 3.2 Philanthropic Givings of Six Companies
During 2007
Corporation
Money Given in 2007
(millions of dollars)
CVS
22.4
Best Buy
31.8
Staples
19.8
Walgreen
9.0
Lowe's
27.5
Wal-Mart
337.9
Notice that the charitable contributions made by Wal-Mart are very large compared to those
of other companies. Hence, it is an outlier. Show how the inclusion of this outlier affects the
value of the mean.
Solution If we do not include the charitable givings of Wal-Mart (the outlier), the mean of
the charitable contributions of the five companies is
22.4 + 31.8 + 19.8 + 9.0 + 27.5 110.5 .
Mean = = — - — = $22.1 million
Now, to see the impact of the outlier on the value of the mean, we include the contributions
of Wal-Mart and find the mean contributions of the six companies. This mean is
22.4 + 31.8 + 19.8 + 9.0 + 27.5 + 337.9 448.4 „
Mean = = = $74.73 million
Thus, including the contributions of Wal-Mart causes more than a threefold increase in the
value of the mean, which changes from $22.1 million to $74.73 million. I
USA TODAY Snapshots®
Packing them in
The Yankees and Mets led the majors in attendance
la st yea rand Figure to draw big crowds again this
season while opening new ballparks. Leading aver-
age attendance figures in 2008:
New York Yankees ^^^^^H 52,585
New York Mets 51,165
Los Angeles Dodgers | 1 46,059
St. Louis Cardinals 42,382
Philadelphia Phillies ^^^H 42,254
Sou rev: & Lis Sports En nrau
By Rent ftubtolebin*, Keucer*
AVERAGE
ATTEN-
DANCE AT
BASEBALL
GAMES
By Mott Yoone jni) Keith Simmons. USA TODAY
The accompanying chart shows five of the Major League Baseball teams that had the highest average
attendance during the 2008 season. According to the information given in the chart, the highest average
,. , , , . Source: USA TODAY, April 2, 2009.
attendance during the 2008 season was for the New York Yankees, which attracted 52,585 spectators copyright © 2009 USA TODAY Chart
per game. reproduced with permission.
The preceding example should encourage us to be cautious. We should remember that the
mean is not always the best measure of central tendency because it is heavily influenced by out-
liers. Sometimes other measures of central tendency give a more accurate impression of a data
set. For example, when a data set has outliers, instead of using the mean, we can use either the
trimmed mean (defined in Exercise 3.33) or the median (to be discussed next) as a measure of
central tendency.
3.1.2 Median
Another important measure of central tendency is the median. It is defined as follows.
Definition
Median The median is the value of the middle term in a data set that has been ranked in increas-
ing order.
As is obvious from the definition of the median, it divides a ranked data set into two equal
parts. The calculation of the median consists of the following two steps:
1. Rank the data set in increasing order.
2. Find the middle term. The value of this term is the median. 1
Note that if the number of observations in a data set is odd, then the median is given by
the value of the middle term in the ranked data. However, if the number of observations is even,
then the median is given by the average of the values of the two middle terms.
'The value of the middle term in a data set ranked in decreasing order will also give the value of the median.
83
84 Chapter 3 Numerical Descriptive Measures
Calculating the median
for ungrouped data: odd
number of data values.
■ EXAMPLE 3-4
The following data give the prices (in thousands of dollars) of seven houses selected from all
houses sold last month in a city.
312 257 421 289 526 374 497
Find the median.
Solution First, we rank the given data in increasing order as follows:
257 289 312 374 421 497 526
Since there are seven homes in this data set and the middle term is the fourth term, the me-
dian is given by the value of the fourth term in the ranked data.
257 289 312 374 421 497 526
t
Median
Thus, the median price of a house is 374, or $374,000. I
f<>>
Calculating the median
for ungrouped data: even
number of data values.
■ EXAMPLE 3-5
Table 3.3 gives the 2008 profits (rounded to billions of dollars) of 12 companies selected from
all over the world.
Table 3.3 Profits of 12 Companies for 2008
Company
2008 Profits
(billions of dollars)
Merck & Co
8
IBM
12
Unilever
7
Microsoft
17
Petrobras
14
Exxon Mobil
45
Lukoil
10
AT&T
13
Nestle
17
Vodafone
13
Deutsche Bank
9
China Mobile
11
Find the median for these data.
Solution First we rank the given profits as follows:
7 8 9 10 11 12 13 13 14 17 17 45
There are 12 values in this data set. Because there is an even number of values in the data set,
the median is given by the average of the two middle values. The two middle values are the
sixth and seventh in the foregoing list of data, and these two values are 12 and 13. The median,
which is given by the average of these two values, is calculated as follows.
7 8 9 10 11 12 13 13 14 17 17 45
t
Median
USA TODAY Snapshots®
The gender pay gap
Among workers 25 or older who
worked full time, year round in
2007, women made an average of
77 cents for every dollar earned by
men. Median earnings:
THE
GENDER
PAY CAP
The accompanying chart shows the median earnings of men and women aged 25 years and older for the
year 2007. These numbers are based on Census Bureau surveys done in 2007 but released in January 2009.
,. , . '. . , , Source: USA TODAY, February 9, 2009.
According to the information given in the chart, in 2007, the median earnings of women 25 years or older copyright © 2009 USA TODAY Chart
were $35,759, whereas those of men 25 years or older were $46,788.
reproduced with permission.
Median
12 + 13 25
12.5 = $12.5 billion
2 2
Thus, the median profit of these 12 companies is $12.5 billion. H
The median gives the center of a histogram, with half of the data values to the left of the
median and half to the right of the median. The advantage of using the median as a measure of
central tendency is that it is not influenced by outliers. Consequently, the median is preferred
over the mean as a measure of central tendency for data sets that contain outliers.
3.1.3 Mode
Mode is a French word that means fashion — an item that is most popular or common. In sta-
tistics, the mode represents the most common value in a data set.
Definition
Mode The mode is the value that occurs with the highest frequency in a data set.
■ EXAMPLE 3-6
The following data give the speeds (in miles per hour) of eight cars that were stopped on
T95 for speeding violations.
77 82 74 81 79 84 74 78
Find the mode.
Solution In this data set, 74 occurs twice, and each of the remaining values occurs only
once. Because 74 occurs with the highest frequency, it is the mode. Therefore,
Calculating the mode for
ungrouped data.
Mode = 74 miles per hour
85
86 Chapter 3 Numerical Descriptive Measures
A major shortcoming of the mode is that a data set may have none or may have more than one
mode, whereas it will have only one mean and only one median. For instance, a data set with each
value occurring only once has no mode. A data set with only one value occurring with the high-
est frequency has only one mode. The data set in this case is called unimodal. A data set with
two values that occur with the same (highest) frequency has two modes. The distribution, in this
case, is said to be bimodal. If more than two values in a data set occur with the same (highest)
frequency, then the data set contains more than two modes and it is said to be multimodal.
Data set with no mode.
■ EXAMPLE 3-7
Last year's incomes of five randomly selected families were $76,150, $95,750, $124,985,
$87,490, and $53,740. Find the mode.
Solution Because each value in this data set occurs only once, this data set contains no
mode. I
Data set with two modes.
■ EXAMPLE 3-8
Refer to the data on 2008 profits of 12 companies given in Table 3.3 of Example 3-5. Find
the mode for these data.
Solution In the data given in Example 3-5, each of the two values 13 and 17 occurs twice,
and each of the remaining values occurs only once. Therefore, that data set has two modes:
$ 1 3 billion and $ 1 7 billion. ■
Data set with three modes.
■ EXAMPLE 3-9
The ages of 10 randomly selected students from a class are 21, 19, 27, 22, 29, 19, 25, 21, 22,
and 30 years, respectively. Find the mode.
Solution This data set has three modes: 19, 21, and 22. Each of these three values occurs
with a (highest) frequency of 2. I
One advantage of the mode is that it can be calculated for both kinds of data — quantitative
and qualitative — whereas the mean and median can be calculated for only quantitative data.
■ EXAMPLE 3-10
The status of five students who are members of the student senate at a college are senior, soph-
omore, senior, junior, and senior, respectively. Find the mode.
Solution Because senior occurs more frequently than the other categories, it is the mode
for this data set. We cannot calculate the mean and median for this data set. I
To sum up, we cannot say for sure which of the three measures of central tendency is a bet-
ter measure overall. Each of them may be better under different situations. Probably the mean
is the most-used measure of central tendency, followed by the median. The mean has the ad-
vantage that its calculation includes each value of the data set. The median is a better measure
when a data set includes outliers. The mode is simple to locate, but it is not of much use in
practical applications.
Finding the mode for
qualitative data.
3.1.4 Relationships Among the Mean, Median, and Mode
As discussed in Chapter 2, two of the many shapes that a histogram or a frequency distribution
curve can assume are symmetric and skewed. This section describes the relationships among
the mean, median, and mode for three such histograms and frequency distribution curves.
3.1 Measures of Central Tendency for Ungrouped Data 87
Knowing the values of the mean, median, and mode can give us some idea about the shape of
a frequency distribution curve.
1. For a symmetric histogram and frequency distribution curve with one peak (see Figure 3.2),
the values of the mean, median, and mode are identical, and they lie at the center of the
distribution.
Figure 3.2 Mean, median, and mode for a symmetric
histogram and frequency distribution curve.
| Variable
Mean = median = mode
2. For a histogram and a frequency distribution curve skewed to the right (see Figure 3.3), the
value of the mean is the largest, that of the mode is the smallest, and the value of the me-
dian lies between these two. (Notice that the mode always occurs at the peak point.) The
value of the mean is the largest in this case because it is sensitive to outliers that occur in
the right tail. These outliers pull the mean to the right.
" Figure 3.3 Mean, median, and mode for a histogram
and frequency distribution curve skewed to the right.
| | \_ ^ Variable
Mode Median Mean
3. If a histogram and a frequency distribution curve are skewed to the left (see Figure 3.4),
the value of the mean is the smallest and that of the mode is the largest, with the value of
the median lying between these two. In this case, the outliers in the left tail pull the mean
to the left.
Figure 3.4 Mean, median, and mode for a histogram
and frequency distribution curve skewed to the left.
7_j t t
Mean Median Mode
Variable
CONCEPTS AND PROCEDURES
3.1 Explain how the value of the median is determined for a data set that contains an odd number of ob-
servations and for a data set that contains an even number of observations.
3.2 Briefly explain the meaning of an outlier. Is the mean or the median a better measure of central ten-
dency for a data set that contains outliers? Illustrate with the help of an example.
3.3 Using an example, show how outliers can affect the value of the mean.
88
Chapter 3 Numerical Descriptive Measures
3.4 Which of the three measures of central tendency (the mean, the median, and the mode) can be cal-
culated for quantitative data only, and which can be calculated for both quantitative and qualitative data?
Illustrate with examples.
3.5 Which of the three measures of central tendency (the mean, the median, and the mode) can assume
more than one value for a data set? Give an example of a data set for which this summary measure as-
sumes more than one value.
3.6 Is it possible for a (quantitative) data set to have no mean, no median, or no mode? Give an exam-
ple of a data set for which this summary measure does not exist.
3.7 Explain the relationships among the mean, median, and mode for symmetric and skewed histograms.
Illustrate these relationships with graphs.
3.8 Prices of cars have a distribution that is skewed to the right with outliers in the right tail. Which of
the measures of central tendency is the best to summarize this data set? Explain.
3.9 The following data set belongs to a population:
5 -7 2 -9 16 10 7
Calculate the mean, median, and mode.
3.10 The following data set belongs to a sample:
14 18 -1 08 8 -16
Calculate the mean, median, and mode.
■ APPLICATIONS
3.11 The following table gives the standard deductions and personal exemptions for persons filing with
"single" status on their 2009 state income taxes in a random sample of 10 states. Calculate the mean and
median for the data on standard deductions for these states.
State
Standard Deduction
(in dollars)
Personal Exemption
(in dollars)
Delaware
3250
110
Hawaii
2000
1040
Kentucky
2100
20
Minnesota
5450
3500
North Dakota
5450
3500
Oregon
1865
169
Rhode Island
5450
3500
Vermont
5450
3500
Virginia
3000
930
Source: TaxFoundation.org.
3.12 Refer to the data table in Exercise 3.1 1. Calculate the mean and median for the data on personal ex-
emptions for these states.
3.13 The following data give the 2007 gross domestic product (GDP) in billions of dollars for all 50 states.
The data are entered in alphabetic order by state (Bureau of Economic Analysis, June 2005).
166
45
247
95
1813
236
216
60
735 397
62
51
610
246
129
117
154
216
48 269
352
382
255
89
229
34
80
127
57 465
76
1103
399
28
466
139
158
531
47 153
34
244
1142
106
25
383
311
58
232 32
a. Calculate the mean and median for these data. Are these values of the mean and the median sam-
ple statistics or population parameters? Explain.
b. Do these data have a mode? Explain.
3.1 Measures of Central Tendency for Ungrouped Data 89
3.14 The following data give the 2008 profits (in millions of dollars) of the six Arizona-based companies
for the year 2008 (Fortune, May 5, 2008). The data represent the following companies, respectively:
Freeport-McMoRan Copper & Gold, Avnet, US Airways Group, Allied Waste Industries, Insight Enter-
prises, and PetSmart.
2977.0 393.1 427.0 273.6 77.8 258.7
Find the mean and median for these data. Do these data have a mode?
3.15 The following data give the 2006-07 team salaries for 20 teams of the English Premier League,
arguably the best-known soccer league in the world. The salaries are given in the order in which the teams
finished during the 2006-07 season. The salaries are in millions of British pounds (note that the approx-
imate value of 1 British pound was $1.95 during the 2006-07 season, so the team salaries range from
$34.3 million to $259 million). (Source: BBC, May 28, 2008.)
92.3
132.8
77.6
89.7
43.8
38.4
30.7
29.8
36.9
36.7
43.2
38.3
62.5
36.4
44.2
35.2
27.5
22.4
34.3
17.6
Find the mean and median for these data.
3.16 The following data give the numbers of car thefts that occurred in a city during the past 12 days.
637 11 4387269 15
Find the mean, median, and mode.
3.17 The following data give the revenues (in millions of dollars) for the last available fiscal year for a
sample of six charitable organizations for serious diseases (Charity Navigator, 2009). The values are, listed
in order, for the Alzheimer's Association, the American Cancer Society, the American Diabetes Associa-
tion, the American Heart Association, the American Lung Association, and the Cystic Fibrosis Foundation.
952 1129 231 668 49 149
Compute the mean and median. Do these data have a mode? Why or why not?
3.18 The following table gives the number of major penalties for each of the 15 teams in the Eastern Con-
ference of the National Hockey League during the 2008-09 season (NHL, 2009). A major penalty is sub-
ject to 5 minutes in the penalty box for a player.
Number of
Team
Major Penalties
Philadelphia
65
Columbus
59
Boston
53
Pittsburgh
51
New York Rangers
50
Tampa Bay
40
Nashville
39
Florida
38
Ottawa
35
Washington
35
Montreal
34
Atlanta
31
New York Islanders
29
Buffalo
26
Toronto
25
Compute the mean and median for the data on major penalties. Do these data have a mode? Why or why not?
3.19 Due to antiquated equipment and frequent windstorms, the town of Oak City often suffers power
outages. The following data give the numbers of power outages for each of the past 12 months.
4 57 3 2023 2 124
Compute the mean, median, and mode for these data.
90 Chapter 3 Numerical Descriptive Measures
3.20 A brochure from the department of public safety in a northern state recommends that motorists should
carry 12 items (flashlights, blankets, and so forth) in their vehicles for emergency use while driving in
winter. The following data give the number of items out of these 12 that were carried in their vehicles by
15 randomly selected motorists.
537801051 21 07 671 19
Find the mean, median, and mode for these data. Are the values of these summary measures population
parameters or sample statistics? Explain.
3.21 Nixon Corporation manufactures computer monitors. The following data are the numbers of com-
puter monitors produced at the company for a sample of 10 days.
24 32 27 23 35 33 29 40 23 28
Calculate the mean, median, and mode for these data.
3.22 The Tri-City School District has instituted a zero-tolerance policy for students carrying any objects
that could be used as weapons. The following data give the number of students suspended during each of
the past 12 weeks for violating this school policy.
1 59 1 21 17 6 9 1 01 43 6 5
Calculate the mean, median, and mode for these data.
3.23 The following data represent the numbers of tornadoes that touched down during 1950 to 1994 in
the 12 states that had the most tornadoes during this period (Storm Prediction Center, 2009). The data for
these states are given in the following order: CO, FL, IA, IL, KS, LA, MO, MS, NE, OK, SD, TX.
1113 2009 1374 1137 2110 1086 1166 1039 1673 2300 1139 5490
a. Calculate the mean and median for these data.
b. Identify the outlier in this data set. Drop the outlier and recalculate the mean and median. Which
of these two summary measures changes by a larger amount when you drop the outlier?
c. Which is the better summary measure for these data, the mean or the median? Explain.
3.24 The following data set lists the number of women from each of 10 different countries who were on
the Rolex Women's World Golf Rankings Top 25 list as of March 31, 2009. The data, entered in that or-
der, are for the following countries: Australia, Brazil, England, Japan, Korea, Mexico, Norway, Sweden,
Taiwan, and United States.
2 1 1 2 9 1 1 2 2 4
a. Calculate the mean and median for these data.
b. Identify the outlier in this data set. Drop the outlier and recalculate the mean and median. Which
of these two summary measures changes by a larger amount when you drop the outlier?
c. Which is the better summary measure for these data, the mean or the median? Explain.
*3.25 One property of the mean is that if we know the means and sample sizes of two (or more) data sets,
we can calculate the combined mean of both (or all) data sets. The combined mean for two data sets is
calculated by using the formula
_ n\X\ + niXi
Combined mean = x = —
M, + « 2
where /?, and n 2 are the sample sizes of the two data sets and X\ and x 2 are the means of the two data sets,
respectively. Suppose a sample of 10 statistics books gave a mean price of $140 and a sample of 8 math-
ematics books gave a mean price of $160. Find the combined mean. (Hint: For this example:
n, = 10, n 2 = 8,3c, = $140, 3c 2 = $160.)
*3.26 Twenty business majors and 18 economics majors go bowling. Each student bowls one game. The
scorekeeper announces that the mean score for the 18 economics majors is 144 and the mean score for
the entire group of 38 students is 150. Find the mean score for the 20 business majors.
*3.27 For any data, the sum of all values is equal to the product of the sample size and mean; that is,
~%x = nx. Suppose the average amount of money spent on shopping by 10 persons during a given week
is $105.50. Find the total amount of money spent on shopping by these 10 persons.
*3.28 The mean 2009 income for five families was $99,520. What was the total 2009 income of these
five families?
*3.29 The mean age of six persons is 46 years. The ages of five of these six persons are 57, 39, 44, 51,
and 37 years, respectively. Find the age of the sixth person.
3.1 Measures of Central Tendency for Ungrouped Data 91
*3.30 Seven airline passengers in economy class on the same flight paid an average of $361 per ticket.
Because the tickets were purchased at different times and from different sources, the prices varied. The
first five passengers paid $420, $210, $333, $695, and $485. The sixth and seventh tickets were purchased
by a couple who paid identical fares. What price did each of them pay?
*3.31 Consider the following two data sets.
Data Set I: 12 25 37 8 41
Data Set II: 19 32 44 15 48
Notice that each value of the second data set is obtained by adding 7 to the corresponding value of the
first data set. Calculate the mean for each of these two data sets. Comment on the relationship between
the two means.
*3.32 Consider the following two data sets.
Data Set I: 4 8 15 9 11
Data Set II: 8 16 30 18 22
Notice that each value of the second data set is obtained by multiplying the corresponding value of the
first data set by 2. Calculate the mean for each of these two data sets. Comment on the relationship be-
tween the two means.
*3.33 The trimmed mean is calculated by dropping a certain percentage of values from each end of a
ranked data set. The trimmed mean is especially useful as a measure of central tendency when a data set
contains a few outliers at each end. Suppose the following data give the ages (in years) of 10 employees
of a company:
47 53 38 26 39 49 19 67 31 23
To calculate the 10% trimmed mean, first rank these data values in increasing order; then drop 10% of the
smallest values and 10% of the largest values. The mean of the remaining 80% of the values will give the
10% trimmed mean. Note that this data set contains 10 values, and 10% of 10 is 1. Thus, if we drop
the smallest value and the largest value from this data set, the mean of the remaining 8 values will be
called the 10% trimmed mean. Calculate the 10% trimmed mean for this data set.
*3.34 The following data give the prices (in thousands of dollars) of 20 houses sold recently in a city.
184 297 365 309 245 387 369 438 195 390
323 578 410 679 307 271 457 795 259 590
Find the 20% trimmed mean for this data set.
*3.35 In some applications, certain values in a data set may be considered more important than others.
For example, to determine students' grades in a course, an instructor may assign a weight to the final exam
that is twice as much as that to each of the other exams. In such cases, it is more appropriate to use the
weighted mean. In general, for a sequence of n data values x l5 x 2 ,..., x n that are assigned weights w u
w 2 ,..., w,„ respectively, the weighted mean is found by the formula
%xw
Weighted mean = . . —
zw
where Xxw is obtained by multiplying each data value by its weight and then adding the products. Sup-
pose an instructor gives two exams and a final, assigning the final exam a weight twice that of each of the
other exams. Find the weighted mean for a student who scores 73 and 67 on the first two exams and 85
on the final. (Hint: Here, x x = 73, x 2 = 67, x 3 = 85, w l = w 2 = 1, and w } = 2.)
*3.36 When studying phenomena such as inflation or population changes that involve periodic increases
or decreases, the geometric mean is used to find the average change over the entire period under study.
To calculate the geometric mean of a sequence of n values x x , x 2 ,..., x n , we multiply them together and
then find the nth root of this product. Thus
Geometric mean = VJq • x 2 • x 3 • ... • x„
Suppose that the inflation rates for the last five years are 4%, 3%, 5%, 6%, and 8%, respectively. Thus at
the end of the first year, the price index will be 1.04 times the price index at the beginning of the year,
and so on. Find the mean rate of inflation over the 5-year period by finding the geometric mean of the
data set 1.04, 1.03, 1.05, 1.06, and 1.08. (Hint: Here, n = 5, % = 1.04, x 2 = 1.03, and so on. Use the x y "
key on your calculator to find the fifth root. Note that the mean inflation rate will be obtained by sub-
tracting 1 from the geometric mean.)
Chapter 3 Numerical Descriptive Measures
3.2 Measures of Dispersion for Ungrouped Data
The measures of central tendency, such as the mean, median, and mode, do not reveal the whole
picture of the distribution of a data set. Two data sets with the same mean may have completely
different spreads. The variation among the values of observations for one data set may be much
larger or smaller than for the other data set. (Note that the words dispersion, spread, and vari-
ation have the same meaning.) Consider the following two data sets on the ages (in years) of
all workers working for each of two small companies.
Company 1: 47 38 35 40 36 45 39
Company 2: 70 33 18 52 27
The mean age of workers in both these companies is the same, 40 years. If we do not know
the ages of individual workers at these two companies and are told only that the mean age of
the workers at both companies is the same, we may deduce that the workers at these two com-
panies have a similar age distribution. As we can observe, however, the variation in the workers'
ages for each of these two companies is very different. As illustrated in the diagram, the ages
of the workers at the second company have a much larger variation than the ages of the workers
at the first company.
Company 1 36 39
LJ
35 38 40 45 47
Company 2
I I I I I
18 27 33 52 70
Thus, the mean, median, or mode by itself is usually not a sufficient measure to reveal the
shape of the distribution of a data set. We also need a measure that can provide some informa-
tion about the variation among data values. The measures that help us learn about the spread of
a data set are called the measures of dispersion. The measures of central tendency and disper-
sion taken together give a better picture of a data set than the measures of central tendency alone.
This section discusses three measures of dispersion: range, variance, and standard deviation.
3.2.1 Range
The range is the simplest measure of dispersion to calculate. It is obtained by taking the dif-
ference between the largest and the smallest values in a data set.
Finding the Range for Ungrouped Data
Range = Largest value — Smallest value
■ EXAMPLE 3-11
Table 3.4 gives the total areas in square miles of the four western South-Central states of the
Calculating the ranee T . . , _
, , United States.
for ungrouped data.
Table 3.4
Total Area
State
(square miles)
Arkansas
53,182
Louisiana
49,651
Oklahoma
69,903
Texas
267,277
Find the range for this data set.
3.2 Measures of Dispersion for Ungrouped Data 93
Solution The maximum total area for a state in this data set is 267,277 square miles, and
the smallest area is 49,651 square miles. Therefore,
Range = Largest value — Smallest value
= 267,277 - 49,651 = 217,626 square miles
Thus, the total areas of these four states are spread over a range of 217,626 square miles. H
The range, like the mean, has the disadvantage of being influenced by outliers. In Example
3-11, if the state of Texas with a total area of 267,277 square miles is dropped, the range de-
creases from 217,626 square miles to 20,252 square miles. Consequently, the range is not a
good measure of dispersion to use for a data set that contains outliers.
Another disadvantage of using the range as a measure of dispersion is that its calculation is
based on two values only: the largest and the smallest. All other values in a data set are ignored
when calculating the range. Thus, the range is not a very satisfactory measure of dispersion.
3.2.2 Variance and Standard Deviation
The standard deviation is the most-used measure of dispersion. The value of the standard de-
viation tells how closely the values of a data set are clustered around the mean. In general, a
lower value of the standard deviation for a data set indicates that the values of that data set are
spread over a relatively smaller range around the mean. In contrast, a larger value of the stan-
dard deviation for a data set indicates that the values of that data set are spread over a relatively
larger range around the mean.
The standard deviation is obtained by taking the positive square root of the variance. The vari-
ance calculated for population data is denoted by tr 2 (read as sigma squared), 2 and the variance
calculated for sample data is denoted by s~. Consequently, the standard deviation calculated for pop-
ulation data is denoted by tr, and the standard deviation calculated for sample data is denoted by
s. Following are what we will call the basic formulas that are used to calculate the variance: 3
%(x - (xf X(x - xf
o~ = and s~ =
N n - 1
where a 2 is the population variance and s 2 is the sample variance.
The quantity x — /jl or x — x in the above formulas is called the deviation of the x value
from the mean. The sum of the deviations of the x values from the mean is always zero; that
is, t(x — fjb) = and %(x — x) = 0.
For example, suppose the midterm scores of a sample of four students are 82, 95, 67, and
92, respectively. Then, the mean score for these four students is
82 + 95 + 67 + 92
x = = 84
4
The deviations of the four scores from the mean are calculated in Table 3.5. As we can observe from
the table, the sum of the deviations of the x values from the mean is zero; that is, %(x — x) = 0.
For this reason we square the deviations to calculate the variance and standard deviation.
Table 3.5
X
x — X
82
82
- 84 =
-2
95
95
- 84 =
+ 11
67
67
- 84 =
-17
92
92
- 84 =
+ 8
X(x -x) =
2 Note that % is uppercase sigma and a is lowercase sigma of the Greek alphabet.
3 From the formula for a 2 , it can be stated that the population variance is the mean of the squared deviations of x values
from the mean. However, this is not true for the variance calculated for a sample data set.
94 Chapter 3 Numerical Descriptive Measures
From the computational point of view, it is easier and more efficient to use short-cut for-
mulas to calculate the variance and standard deviation. By using the short-cut formulas, we
reduce the computation time and round-off errors. Use of the basic formulas for ungrouped data
is illustrated in Section A3. 1.1 of Appendix 3.1 of this chapter. The short-cut formulas for
calculating the variance and standard deviation are given next.
Short-Cut Formulas for the Variance and Standard Deviation for Ungrouped Data
o-- = and s =
N n-\
where a 1 is the population variance and s 2 is the sample variance.
The standard deviation is obtained by taking the positive square root of the variance.
^2
Population standard deviation: a = V o"
Sample standard deviation:
Note that the denominator in the formula for the population variance is N, but that in the for-
mula for the sample variance it is n — 1 . 4
to billions of dollars) of five
rket Value
ns of dollars)
75
107
271
138
71
Solution Let x denote the 2008 market value (in billions of dollars) of a company. The values
of Xx and Xx 2 are calculated in Table 3.6.
Table 3.6
X
x 2
75
5625
107
11,449
271
73,441
138
19,044
71
5041
Xx = 662
Xx 2 = 114,600
Calculation of the variance involves the following four steps.
Calculating the sample variance
and standard deviation
for ungrouped data.
■ EXAMPLE 3-12
The following table gives the 2008 market values (rounc
international companies.
Company
PepsiCo
Google
PetroChina
Johnson & Johnson
Intel
Mai
(billioi
Find the variance and standard deviation for these data.
4 The reason that the denominator in the sample formula is n — 1 and not n follows: The sample variance underesti-
mates the population variance when the denominator in the sample formula for variance is n. However, the sample
variance does not underestimate the population variance if the denominator in the sample formula for variance is
n — 1 . In Chapter 8 we will learn that n — 1 is called the degrees of freedom.
3.2 Measures of Dispersion for Ungrouped Data 95
Step 1. Calculate Xx.
The sum of the values in the first column of Table 3.6 gives the value of Xx, which
is 662.
Step 2. Find Xx 2 .
The value of Xx 2 is obtained by squaring each value of x and then adding the squared val-
ues. The results of this step are shown in the second column of Table 3.6. Notice that
Xx 2 = 1 14,600.
Step 3. Determine the variance.
Substitute all the values in the variance formula and simplify. Because the given data
are on the market values of only five companies, we use the formula for the sample
variance.
(Sx) 2 (662) 2
2 ^~^r 114 ' 600 -^ 114,600 - 87,648.80
s 2 = = = = 6737.80
n - 1 5-1 4
Step 4. Obtain the standard deviation.
The standard deviation is obtained by taking the (positive) square root of the variance.
s = V6737.80 = 82.0841 = $82.08 billion
Thus, the standard deviation of the market values of these five companies is $82.08
billion.
1. The values of the variance and the standard deviation are never negative. That is, Two Observations
the numerator in the formula for the variance should never produce a negative value.
Usually the values of the variance and standard deviation are positive, but if a data set
has no variation, then the variance and standard deviation are both zero. For example,
if four persons in a group are the same age — say, 35 years — then the four values in the
data set are
35 35 35 35
If we calculate the variance and standard deviation for these data, their values are zero. This
is because there is no variation in the values of this data set.
2. The measurement units of variance are always the square of the measurement units
of the original data. This is so because the original values are squared to calculate the
variance. In Example 3-12, the measurement units of the original data are billions of dol-
lars. However, the measurement units of the variance are squared billions of dollars,
which, of course, does not make any sense. Thus, the variance of the 2008 market val-
ues of these five companies in Example 3-12 is 6737.80 squared billion dollars. But the
measurement units of the standard deviation are the same as the measurement units of
the original data because the standard deviation is obtained by taking the square root of
the variance.
EXAMPLE 3-13
Following are the 2009 earnings (in thousands of dollars) before taxes for all six employees
of a small company.
88.50 108.40 65.50 52.50 79.80 54.60
Calculate the variance and standard deviation for these data.
Calculating the population
variance and standard deviation
for ungrouped data.
Solution Let x denote the 2009 earnings before taxes of an employee of this company. The
values of Xx and Xx 2 are calculated in Table 3.7.
96 Chapter 3 Numerical Descriptive Measures
Table 3.7
X
x 2
00. jU
ion tc
IooZ.Zj
108.40
11,750.56
65.50
4290.25
52.50
2756.25
79.80
6368.04
54.60
2981.16
= 449.30
2x 2 = 35,978.51
Because the data in this example are on earnings of all employees of this company, we use
the population formula to compute the variance. Thus, the variance is
(Xx) 2 (449.30) 2
Xx 2 - 35,978.51 - -
6 — = 388.90
N 6
The standard deviation is obtained by taking the (positive) square root of the variance:
a = V388.90 = 19.721 thousand = $19,721
Thus, the standard deviation of the 2009 earnings of all six employees of this company is
$19,721. ■
Warning ► Note that Xx 2 is not the same as (Xx) 2 . The value of Xx 2 is obtained by squaring the x values
and then adding them. The value of (Xx) 2 is obtained by squaring the value of Xx.
The uses of the standard deviation are discussed in Section 3.4. Later chapters explain how
the mean and the standard deviation taken together can help in making inferences about the
population.
3.2.3 Population Parameters and Sample Statistics
A numerical measure such as the mean, median, mode, range, variance, or standard deviation
calculated for a population data set is called a population parameter, or simply a parameter.
A summary measure calculated for a sample data set is called a sample statistic, or simply a
statistic. Thus, /j, and <x are population parameters, and x and s are sample statistics. As an il-
lustration, x = $228 billion in Example 3-1 is a sample statistic, and j± = 45.25 years in Ex-
ample 3-2 is a population parameter. Similarly, s = $82.08 billion in Example 3-12 is a sam-
ple statistic, whereas a = $19,721 in Example 3-13 is a population parameter.
EXERCISES
CONCEPTS AND PROCEDURES
3.37 The range, as a measure of spread, has the disadvantage of being influenced by outliers. Illustrate
this with an example.
3.38 Can the standard deviation have a negative value? Explain.
3.39 When is the value of the standard deviation for a data set zero? Give one example. Calculate the
standard deviation for the example and show that its value is zero.
3.40 Briefly explain the difference between a population parameter and a sample statistic. Give one ex-
ample of each.
3.41 The following data set belongs to a population:
5 -7 2 -9 1 61 07
Calculate the range, variance, and standard deviation.
3.2 Measures of Dispersion for Ungrouped Data 97
3.42 The following data set belongs to a sample:
14 18 -1 08 8 -16
Calculate the range, variance, and standard deviation.
■ APPLICATIONS
3.43 The following data give the number of shoplifters apprehended during each of the past 8 weeks at a
large department store.
7 1 08 3 1 51 26 1 1
a. Find the mean for these data. Calculate the deviations of the data values from the mean. Is the
sum of these deviations zero?
b. Calculate the range, variance, and standard deviation.
3.44 The following data give the prices of seven textbooks randomly selected from a university bookstore.
$89 $170 $104 $113 $56 $161 $147
a. Find the mean for these data. Calculate the deviations of the data values from the mean. Is the
sum of these deviations zero?
b. Calculate the range, variance, and standard deviation.
3.45 The following data give the numbers of car thefts that occurred in a city in the past 12 days.
6371 14 38726915
Calculate the range, variance, and standard deviation.
3.46 Refer to the data in Exercise 3.23, which contained the numbers of tornadoes that touched down in
12 states that had the most tornadoes during the period 1950 to 1994. The data are reproduced here.
1113 2009 1374 1137 2110 1086 1166 1039 1673 2300 1139 5490
Find the variance, standard deviation, and range for these data.
3.47 The following data give the numbers of pieces of junk mail received by 10 families during the past
month.
41 33 28 21 29 19 14 31 39 36
Find the range, variance, and standard deviation.
3.48 The following data give the number of highway collisions with large wild animals, such as deer or
moose, in one of the northeastern states during each week of a 9-week period.
71 03 825749
Find the range, variance, and standard deviation.
3.49 Attacks by stinging insects, such as bees or wasps, may become medical emergencies if either the
victim is allergic to venom or multiple stings are involved. The following data give the number of patients
treated each week for such stings in a large regional hospital during 13 weeks last summer.
1523041701201
Compute the range, variance, and standard deviation for these data.
3.50 The following data give the number of hot dogs consumed by 10 participants in a hot-dog-eating
contest.
21 17 32 8 20 15 17 23 9 18
Calculate the range, variance, and standard deviation for these data.
3.51 Following are the temperatures (in degrees Fahrenheit) observed during eight wintry days in a mid-
western city:
23 14 6 -7 -2 11 16 19
Compute the range, variance, and standard deviation.
3.52 The following data give the numbers of hours spent partying by 10 randomly selected college stu-
dents during the past week.
7 1 45 9 7 1 04 8
Compute the range, variance, and standard deviation.
98
Chapter 3 Numerical Descriptive Measures
3.53 The following data represent the total points scored in each of the NFL championship games played
from 2000 through 2009 in that order.
39 41 37 69 61 45 31 46 31 50
Compute the variance, standard deviation, and range for these data.
3.54 The following data represent the 2006 guaranteed annual salaries (in thousands of dollars) of the head
coaches of the final eight teams in the 2006 NCAA Men's Basketball Championship. The data are given in
the following order: Connecticut, Florida, George Mason, LSU, Memphis, Texas, UCLA, and Villanova.
1500 1389 489 900 1315 1800 1150 584
Compute the variance, standard deviation, and range for these data.
3.55 The following data give the hourly wage rates of eight employees of a company.
$22 22 22 22 22 22 22 22
Calculate the standard deviation. Is its value zero? If yes, why?
3.56 The following data are the ages (in years) of six students.
19 19 19 19 19 19
Calculate the standard deviation. Is its value zero? If yes, why?
*3.57 One disadvantage of the standard deviation as a measure of dispersion is that it is a measure of ab-
solute variability and not of relative variability. Sometimes we may need to compare the variability of two
different data sets that have different units of measurement. The coefficient of variation is one such meas-
ure. The coefficient of variation, denoted by CV, expresses standard deviation as a percentage of the mean
and is computed as follows:
For population data: CV = — X 100%
M
For sample data: CV = = X 100%
x
The yearly salaries of all employees who work for a company have a mean of $62,350 and a standard de-
viation of $6820. The years of experience for the same employees have a mean of 15 years and a stan-
dard deviation of 2 years. Is the relative variation in the salaries larger or smaller than that in years of ex-
perience for these employees?
*3.58 The SAT scores of 100 students have a mean of 975 and a standard deviation of 105. The CPAs of
the same 100 students have a mean of 3.16 and a standard deviation of .22. Is the relative variation in SAT
scores larger or smaller than that in CPAs?
*3.59 Consider the following two data sets.
Data Set I: 12 25 37 8 41
Data Set II: 19 32 44 15 48
Note that each value of the second data set is obtained by adding 7 to the corresponding value of the first
data set. Calculate the standard deviation for each of these two data sets using the formula for sample data.
Comment on the relationship between the two standard deviations.
*3.60 Consider the following two data sets.
Data Set I: 4 8 15 9 11
Data Set II: 8 16 30 18 22
Note that each value of the second data set is obtained by multiplying the corresponding value of the first
data set by 2. Calculate the standard deviation for each of these two data sets using the formula for pop-
ulation data. Comment on the relationship between the two standard deviations.
3.3 Mean, Variance, and Standard Deviation
for Grouped Data
In Sections 3.1.1 and 3.2.2, we learned how to calculate the mean, variance, and standard de-
viation for ungrouped data. In this section, we will learn how to calculate the mean, variance,
and standard deviation for grouped data.
3.3 Mean, Variance, and Standard Deviation for Grouped Data 99
3.3.1 Mean for Grouped Data
We learned in Section 3.1.1 that the mean is obtained by dividing the sum of all values by the
number of values in a data set. However, if the data are given in the form of a frequency table,
we no longer know the values of individual observations. Consequently, in such cases, we can-
not obtain the sum of individual values. We find an approximation for the sum of these values
using the procedure explained in the next paragraph and example. The formulas used to calcu-
late the mean for grouped data follow.
Calculating Mean for Grouped Data
Mean for population data: jj.
Mean for sample data: x
where m is the midpoint and /is the frequency of a class.
To calculate the mean for grouped data, first find the midpoint of each class and then mul-
tiply the midpoints by the frequencies of the corresponding classes. The sum of these products,
denoted by %mf, gives an approximation for the sum of all values. To find the value of the mean,
divide this sum by the total number of observations in the data.
■ EXAMPLE 3-14
Table 3.8 gives the frequency distribution of the daily commuting times (in minutes) from
home to work for all 25 employees of a company.
Table 3.8
Daily Commuting Time Number of
(minutes) Employees
to less than 10 4
10 to less than 20 9
20 to less than 30 6
30 to less than 40 4
40 to less than 50 2
Calculate the mean of the daily commuting times.
Solution Note that because the data set includes all 25 employees of the company, it rep-
resents the population. Table 3.9 shows the calculation of %mf. Note that in Table 3.9, m
denotes the midpoints of the classes.
Table 3.9
Daily Commuting Time
(minutes)
/
in
mf
to less than 10
4
5
20
10 to less than 20
9
15
135
20 to less than 30
6
25
150
30 to less than 40
4
35
140
40 to less than 50
2
45
90
N = 25
tmf = 535
Zrrtf
N
Calculating the population
mean for grouped data.
100 Chapter 3 Numerical Descriptive Measures
To calculate the mean, we first find the midpoint of each class. The class midpoints are
recorded in the third column of Table 3.9. The products of the midpoints and the correspon-
ding frequencies are listed in the fourth column. The sum of the fourth column values, de-
noted by %mf, gives the approximate total daily commuting time (in minutes) for all 25 em-
ployees. The mean is obtained by dividing this sum by the total frequency. Therefore,
%mf 535 „, An .
ii = = = 21.40 minutes
Af 25
Thus, the employees of this company spend an average of 21.40 minutes a day commuting
from home to work.
What do the numbers 20, 135, 150, 140, and 90 in the column labeled mf in Table 3.9 rep-
resent? We know from this table that 4 employees spend to less than 10 minutes commuting
per day. If we assume that the time spent commuting by these 4 employees is evenly spread in
the interval to less than 10, then the midpoint of this class (which is 5) gives the mean time
spent commuting by these 4 employees. Hence, 4 X 5 = 20 is the approximate total time (in
minutes) spent commuting per day by these 4 employees. Similarly, 9 employees spend 10 to
less than 20 minutes commuting per day, and the total time spent commuting by these 9 em-
ployees is approximately 135 minutes a day. The other numbers in this column can be inter-
preted in the same way. Note that these numbers give the approximate commuting times for
these employees based on the assumption of an even spread within classes. The total commut-
ing time for all 25 employees is approximately 535 minutes. Consequently, 21.40 minutes is an
approximate and not the exact value of the mean. We can find the exact value of the mean only
if we know the exact commuting time for each of the 25 employees of the company.
Calculating the sample mean
for grouped data.
■ EXAMPLE 3-15
Table 3.10 gives the frequency distribution of the number of orders received each day during
the past 50 days at the office of a mail-order company.
Table 3.10
Number of Orders
Number of Days
10-12
4
13-15
12
16-18
20
19-21
14
Calculate the mean.
Solution Because the data set includes only 50 days, it represents a sample. The value of
Xmf is calculated in Table 3.11.
Table 3.11
Number of Orders
/
m
mf
10-12
4
11
44
13-15
12
14
168
16-18
20
17
340
19-21
14
20
280
n = 50
2m/ = 832
The value of the sample mean is
_ Xmf 832
16.64 orders
50
3.3 Mean, Variance, and Standard Deviation for Grouped Data 101
Thus, this mail-order company received an average of 16.64 orders per day during these
50 days. ■
3.3.2 Variance and Standard Deviation for Grouped Data
Following are what we will call the basic formulas used to calculate the population and sample
variances for grouped data:
X/(m-M) 2 X/(m-I) 2
a = and s
V n - 1
where a 2 is the population variance, s 2 is the sample variance, and m is the midpoint of a class.
In either case, the standard deviation is obtained by taking the positive square root of the
variance.
Again, the short-cut formulas are more efficient for calculating the variance and standard
deviation. Section A3. 1.2 of Appendix 3.1 at the end of this chapter shows how to use the ba-
sic formulas to calculate the variance and standard deviation for grouped data.
Short-Cut Formulas for the Variance and Standard Deviation for Grouped Data
- 2 , (W) 2 „ 2f (Xm/) 2
Zm f Xm f ■
N n
and s =
N n - 1
a 2 = and s 2
where a 2 is the population variance, s 2 is the sample variance, and m is the midpoint of a class.
The standard deviation is obtained by taking the positive square root of the variance.
Population standard deviation: cr -
Sample standard deviation: j = Vf
2
Examples 3-16 and 3-17 illustrate the use of these formulas to calculate the variance and
standard deviation.
EXAMPLE 3-16
The following data, reproduced from Table 3.8 of Example 3-14, give the frequency distri-
bution of the daily commuting times (in minutes) from home to work for all 25 employees of . , _ , , , . ^
J ° variance and standard deviation
a company. for grouped data.
Daily Commuting Time
(minutes)
Number of Employees
to less than 10
4
10 to less than 20
9
20 to less than 30
6
30 to less than 40
4
40 to less than 50
2
Calculate the variance and standard deviation.
Solution All four steps needed to calculate the variance and standard deviation for grouped
data are shown after Table 3.12.
102
Chapter 3 Numerical Descriptive Measures
Table 3.12
Daily Commuting Time
(minutes)
/
m
mf
m 2 f
to less than 10
4
5
20
100
10 to less than 20
9
15
135
2025
20 to less than 30
6
25
150
3750
30 to less than 40
4
35
140
4900
40 to less than 50
2
45
90
4050
N — 25
2m/ = 535
2m 2 / = 14,825
Step 1. Calculate the value of %mf.
To calculate the value of Xmf first find the midpoint m of each class (see the third column
in Table 3.12) and then multiply the corresponding class midpoints and class frequencies
(see the fourth column). The value of 2m/ is obtained by adding these products. Thus,
2m/ = 535
Step 2. Find the value of Xm f.
To find the value of 2m 2 /, square each m value and multiply this squared value of m by
the corresponding frequency (see the fifth column in Table 3.12). The sum of these prod-
ucts (that is, the sum of the fifth column) gives 2m 2 /. Hence,
2m 2 / = 14,825
Step 3. Calculate the variance.
Because the data set includes all 25 employees of the company, it represents the popula-
tion. Therefore, we use the formula for the population variance:
, (2m/) 2 (535) 2
2m 2 / - — 14,825 -
25 3376 135.04
N 25 25
Step 4. Calculate the standard deviation.
To obtain the standard deviation, take the (positive) square root of the variance.
o- = Vct 2 = V135.04 = 11.62 minutes
Thus, the standard deviation of the daily commuting times for these employees is 11.62
minutes. H
Note that the values of the variance and standard deviation calculated in Example 3-16 for
grouped data are approximations. The exact values of the variance and standard deviation can be
obtained only by using the ungrouped data on the daily commuting times of the 25 employees.
Calculating the sample
variance and standard
deviation for grouped data.
■ EXAMPLE 3-17
The following data, reproduced from Table 3.10 of Example 3-15, give the frequency distri-
bution of the number of orders received each day during the past 50 days at the office of a
mail-order company.
Number of Orders
/
10-12
4
13-15
12
16-18
20
19-21
14
Calculate the variance and standard deviation.
3.3 Mean, Variance, and Standard Deviation for Grouped Data 103
Solution All the information required for the calculation of the variance and standard de-
viation appears in Table 3.13.
Table 3.13
Number of Orders
/
in
mf
m 2 f
10-12
4
11
44
484
13-15
12
14
168
2352
16-18
20
17
340
5780
19-21
14
20
280
5600
n = 50
2m/ = 832
2m 2 / = 14,216
Because the data set includes only 50 days, it represents a sample. Hence, we use the sam-
ple formulas to calculate the variance and standard deviation. By substituting the values into
the formula for the sample variance, we obtain
(2m/) 2 (832) 2
Sm 2 / 14,216 —
n TX 50
= 7.5820
n - 1 50-1
Hence, the standard deviation is
s = V? = V7.5820 = 2.75 orders
Thus, the standard deviation of the number of orders received at the office of this mail-order
company during the past 50 days is 2.75.
EXERCISES
CONCEPTS AND PROCEDURES
3.61 Are the values of the mean and standard deviation that are calculated using grouped data exact or
approximate values of the mean and standard deviation, respectively? Explain.
3.62 Using the population formulas, calculate the mean, variance, and standard deviation for the follow-
ing grouped data.
X
2-4
5-7
8-10
11-13
14-16
f
5
9
1
4 7
5
3.63 Using the sample formulas, find the mean, variance, and standard deviation for the grouped data dis-
played in the following table.
x /
to less than 4 17
4 to less than 8 23
8 to less than 12 15
12 to less than 16 11
16 to less than 20 8
20 to less than 24 6
■ APPLICATIONS
3.64 The following table gives the frequency distribution of the amounts of telephone bills for October
2009 for a sample of 50 families.
104 Chapter 3 Numerical Descriptive Measures
A t r« rri I l it'll
Amount of Telephone Bill
I tit Jlltil a ^
ii iiiiiuci ui i amines
40 to less man /0
9
70 to less than 100
11
100 to less than 130
16
130 to less than 160
10
160 to less than 190
4
Calculate the mean, variance, and standard deviation.
3.65 The following table gives the frequency distribution of the number of hours spent per week playing
video games by all 60 students of the eighth grade at a school.
Hours per Week
Number of Students
to less than 5
7
5 to less than 10
12
10 to less than 15
15
15 to less than 20
13
20 to less than 25
8
25 to less than 30
5
Find the mean, variance, and standard deviation.
3.66 The following table gives the grouped data on the weights of all 100 babies born at a hospital in
2009.
Weight (pounds) Number of Babies
3 to less than 5 5
5 to less than 7 30
7 to less than 9 40
9 to less than 1 1 20
1 1 to less than 13 5
Find the mean, variance, and standard deviation.
3.67 The following table gives the frequency distribution of the total miles driven during 2009 by 300 car
owners.
Miles Driven in 2009
(in thousands) Number of Car Owners
to less than 5
7
5 to less than 10
26
10 to less than 15
59
15 to less than 20
71
20 to less than 25
62
25 to less than 30
39
30 to less than 35
22
35 to less than 40
14
Find the mean, variance, and standard deviation. Give a brief interpretation of the values in the column
labeled mf'm your table of calculations. What does 2m/ represent?
3.68 The following table gives information on the amounts (in dollars) of electric bills for August 2009
for a sample of 50 families.
3.4 Use of Standard Deviation
105
A j_ J* T711 j • it 'II
Amount ot Electric Bill
11 lllllUCI 111 J. <III1111L>>
to less than 40
5
40 to less than 80
16
80 to less than 120
11
120 to less than 160
10
160 to less than 200
8
Find the mean, variance, and standard deviation. Give a brief interpretation of the values in the column
labeled m/in your table of calculations. What does represent?
3.69 For 50 airplanes that arrived late at an airport during a week, the time by which they were late was
observed. In the following table, x denotes the time (in minutes) by which an airplane was late, and /de-
notes the number of airplanes.
X
/
to less than 20
14
20 to less than 40
18
40 to less than 60
9
60 to less than 80
5
80 to less than 100
4
Find the mean, variance, and standard deviation.
3.70 The following table gives the frequency distribution of the number of errors committed by a college
baseball team in all of the 45 games that it played during the 2008-09 season.
Number of Errors
Number of Games
11
1
14
2
9
3
7
4
3
5
1
Find the mean, variance, and standard deviation. (Hint: The classes in this example are single valued. These
values of classes will be used as values of m in the formulas for the mean, variance, and standard deviation.)
3.71 Spot prices per barrel of crude oil reached their highest levels in history during June and July of
2008. The following data give the spot prices (in dollars) of a barrel of crude oil for 14 business days from
June 30, 2008, through July 18, 2008 (Energy Information Administration, April 15, 2009).
139.96 141.06 143.74 145.31 141.38 136.06 135.88
141.47 144.96 145.16 138.68 134.63 129.43 128.94
a. Find the mean for these data.
b. Construct a frequency distribution table for these data using a class width of 3.00 and the lower
boundary of the first class equal to 128.00.
c. Using the method of Section 3.3.1, find the mean of the grouped data of part b.
d. Compare your means from parts a and c. If the two means are not equal, explain why they differ.
3.4 Use of Standard Deviation
By using the mean and standard deviation, we can find the proportion or percentage of the to-
tal observations that fall within a given interval about the mean. This section briefly discusses
Chebyshev's theorem and the empirical rule, both of which demonstrate this use of the stan-
dard deviation.
106 Chapter 3 Numerical Descriptive Measures
3.4.1 Chebyshev's Theorem
Chebyshev's theorem gives a lower bound for the area under a curve between two points that
are on opposite sides of the mean and at the same distance from the mean.
Definition
Chebyshev's Theorem For any number k greater than 1, at least (1 — l/k 2 ) of the data values lie
within k standard deviations of the mean.
Figure 3.5 illustrates Chebyshev's theorem.
1
- = 1 - .25 = .75or75 £
Thus, for example, if k = 2, then
_ j_ _ 1
Therefore, according to Chebyshev's theorem, at least .75, or 75%, of the values of a data set
lie within two standard deviations of the mean. This is shown in Figure 3.6.
Figure 3.6 Percentage of values within two standard
deviations of the mean for Chebyshev's theorem.
At least 75% of
the values lie in
the shaded areas
If k = 3, then,
k 2
1
1
p. -2a
1
W 2 =
1
u + 2o
1 - .11 = .89 or 89% approximately
According to Chebyshev's theorem, at least .89, or 89%, of the values fall within three standard
deviations of the mean. This is shown in Figure 3.7.
Figure 3.7 Percentage of values within three
standard deviations of the mean for Chebyshev's
theorem.
At least 89% of
the values lie in
the shaded areas
Although in Figures 3.5 through 3.7 we have used the population notation for the mean and
standard deviation, the theorem applies to both sample and population data. Note that Chebyshev's
theorem is applicable to a distribution of any shape. However, Chebyshev's theorem can be used
3.4 Use of Standard Deviation
107
only for k > 1 . This is so because when k
the value of 1 — 1 /k 2 is negative.
1, the value of 1 — 1/k 2 is zero, and when k < 1,
■ EXAMPLE 3-18
The average systolic blood pressure for 4000 women who were screened for high blood pres-
sure was found to be 187 mm Hg with a standard deviation of 22. Using Chebyshev's theo-
rem, find at least what percentage of women in this group have a systolic blood pressure be-
tween 143 and 231 mm Hg.
Solution Let fx and a be the mean and the standard deviation, respectively, of the systolic
blood pressures of these women. Then, from the given information,
/JL = 187 and cr = 22
To find the percentage of women whose systolic blood pressures are between 143 and
231 mm Hg, the first step is to determine k. As shown below, each of the two points, 143 and
231, is 44 units away from the mean.
Applying Chebyshev's theorem.
143 - 187
-44-
•231 - 187 = 44-
143
fi = 187
231
The value of k is obtained by dividing the distance between the mean and each point by the
standard deviation. Thus,
'k 1
k = 44/22
1 1
W~ 1 ~4 =
= 2
1 - .25
.75 or 75%
At least 75% of the
women have systolic
blood pressure between
143 and 231
231
H +2a
Systolic blood
pressure
Figure 3.8 Percentage of women with systolic blood pressure
between 143 and 231.
Hence, according to Chebyshev's theorem, at least 75% of the women have systolic blood
pressure between 143 and 231 mm Hg. This percentage is shown in Figure 3.8. I
3.4.2 Empirical Rule
Whereas Chebyshev's theorem is applicable to any kind of distribution, the empirical rule
applies only to a specific type of distribution called a bell-shaped distribution, as shown in
Figure 3.9. More will be said about such a distribution in Chapter 6, where it is called a nor-
mal curve. In this section, only the following three rules for the curve are given.
Empirical Rule For a bell-shaped distribution, approximately
1. 68% of the observations lie within one standard deviation of the mean.
2. 95% of the observations lie within two standard deviations of the mean.
3. 99.7% of the observations lie within three standard deviations of the mean.
Figure 3.9 illustrates the empirical rule. Again, the empirical rule applies to both popula-
tion data and sample data.
HERE
COMES
THE SD
Source: Daniel Seligman, "Here comes
the SD," Fortune, May 15, 1995.
Copyright © 1995, The Time Inc.
Reproduced with permission. All rights
reserved.
When your servant first became a Fortune writer several decades ago, it was hard doctrine that "several"
meant three to eight, also that writers must not refer to "gross national product" without pausing to define
this arcane term. GNP was in fact a relatively new concept at the time, having been introduced to the coun-
try only several years previously— in Roosevelt's 1944 budget message— so the presumption that readers had
to be told repeatedly it was the "value of all goods and services produced by the economy" seemed entirely
reasonable to this young writer, who personally had to look up the definition every time.
Numeracy lurches on. Nowadays the big question for editors is whether an average college-educated
bloke needs a handhold when confronted with the term "standard deviation." The SD is suddenly on-
stage because the Securities and Exchange Commission is wondering aloud whether investment com-
panies should be required to tell investors the standard deviation of their mutual funds' total returns
over various past periods. Barry Barbash, SEC director of investment management, favors the require-
ment but confessed to the Washington Post that he worries about investors who will think a standard
deviation is the dividing line on a highway or something.
The view around here is that the SEC is performing a noble service, but only partly because the re-
quirement would enhance folks' insights into mutual funds. The commission's underlying idea is to give
investors a better and more objective measure than is now available of the risk associated with differ-
ent kinds of portfolios. The SD is a measure of variability, and funds with unusually variable returns—
sometimes very high, sometimes very low— are presumed to be more risky.
What one really likes about the proposal, however, is the prospect that it will incentivize millions of
greedy Americans to learn a little elementary statistics. One already has a list of issues that could be dis-
cussed much more thrillingly if only your average liberal arts graduate had a glimmer about the SD and
the normal curve. The bell-shaped normal curve, or rather, the area underneath the curve, shows you
how Providence arranged for things to be distributed in our world— with people's heights, or incomes, or
IQs, or investment returns bunched around middling outcomes, and fewer and fewer cases as you move
down and out toward the extremes. A line down the center of the curve represents the mean outcome,
and deviations from the mean are measured by the SD.
An amazing property of the SD is that exactly 68.26% of all normally distributed data are within one SD
of the mean. We once asked a professor of statistics a question that seemed to us quite profound, to wit,
why that particular figure? The Prof answered dismissively that God had decided on 68.26% for exactly the
same reason He had landed on 3.14 as the ratio between circumferences and diameters— because He just
felt like it. The Almighty has also proclaimed that 95.44% of all data are within two SDs of the mean, and
99.73% within three SDs. When you know the mean and SD of some outcome, you can instantly establish
the percentage probability of its occurrence. White men's heights in the U.S. average 69.2 inches, with an SD
of 2.8 inches (according to the National Center for Health Statistics), which means that a 6-foot-5 chap is in
the 99th percentile. In 1994, scores on the verbal portion of the Scholastic Assessment Test had a mean of
423 and an SD of 113, so if you scored 649— two SDs above the mean— you were in the 95th percentile.
As the SEC is heavily hinting, average outcomes are interesting but for many purposes inadequate; one
also yearns to know the variability around that average. From 1926 through 1994, the S&P 500 had an av-
erage annual return of just about 10%. The SD accompanying that figure was just about 20%. Since returns
will be within 1 SD some 68% of the time, they will be more than 1 SD from the mean 32% of the time.
And since half these swings will be on the downside, we expect fund owners to lose more than 10% of
their money about one year out of six and to lose more than 30% (two SDs below the mean) about one
year out of 20. If your time horizon is short and you can't take losses like that, you arguably don't belong in
stocks. If you think SDs are highway dividers, you arguably don't belong in cars.
Figure 3.9 Illustration of the empirical rule.
(i-3o ji-2o u-o u, |i + u + 2au + 3o
108
3.4 Use of Standard Deviation
109
■ EXAMPLE 3-19
The age distribution of a sample of 5000 persons is bell shaped with a mean of 40 years and \ ~
a standard deviation of 12 years. Determine the approximate percentage of people who are 16 Applying the empirical rule.
to 64 years old.
Solution We use the empirical rule to find the required percentage because the distribution
of ages follows a bell-shaped curve. From the given information, for this distribution,
x = 40 years and s = 12 years
16 x = 40 64 Ages
t t
x - 2s x + 2s
Figure 3.10 Percentage of people who are 16 to 64 years old.
Each of the two points, 16 and 64, is 24 units away from the mean. Dividing 24 by 12, we con-
vert the distance between each of the two points and the mean in terms of standard deviations.
Thus, the distance between 16 and 40 and that between 40 and 64 is each equal to 2s. Conse-
quently, as shown in Figure 3.10, the area from 16 to 64 is the area from 3c — 2s to 3c + 2s.
Because the area within two standard deviations of the mean is approximately 95% for a bell-
shaped curve, approximately 95% of the people in the sample are 16 to 64 years old. I
EXERCISES
CONCEPTS AND PROCEDURES
3.72 Briefly explain Chebyshev's theorem and its applications.
3.73 Briefly explain the empirical rule. To what kind of distribution is it applied?
3.74 A sample of 2000 observations has a mean of 74 and a standard deviation of 12. Using Chebyshev's
theorem, find at least what percentage of the observations fall in the intervals x ± 2s, x ± 2.5s, and
x ± 3s. Note that here x ± 2s represents the interval x — 2s to x + 2s, and so on.
3.75 A large population has a mean of 230 and a standard deviation of 41. Using Chebyshev's theorem,
find at least what percentage of the observations fall in the intervals /x ± 2a, /jl ± 2.5cr, and /jl ± 3cr.
3.76 A large population has a mean of 310 and a standard deviation of 37. Using the empirical rule, find
what percentage of the observations fall in the intervals /x ± la, /jl ± 2a, and /jl ± 3a.
3.77 A sample of 3000 observations has a mean of 82 and a standard deviation of 16. Using the empiri-
cal rule, find what percentage of the observations fall in the intervals x ± Is, x ± 2s, and x ± 3s.
■ APPLICATIONS
3.78 The mean time taken by all participants to run a road race was found to be 220 minutes with a stan-
dard deviation of 20 minutes. Using Chebyshev's theorem, find the percentage of runners who ran this
road race in
a. 180 to 260 minutes b. 160 to 280 minutes c. 170 to 270 minutes
3.79 The 2009 gross sales of all companies in a large city have a mean of $2.3 million and a standard de-
viation of $.6 million. Using Chebyshev's theorem, find at least what percentage of companies in this city
had 2009 gross sales of
a. $1.1 to $3.5 million b. $.8 to $3.8 million c. $.5 to $4.1 million
110
Chapter 3 Numerical Descriptive Measures
3.80 Suppose the average credit card debt for households currently is $9500 with a standard deviation of $2600.
a. Using Chebyshev's theorem, find at least what percentage of current credit card debts for all house-
holds are between
i. $4300 and $14,700 ii. $3000 and $16,000
*b. Using Chebyshev's theorem, find the interval that contains credit card debts of at least 89% of all
households.
3.81 The mean monthly mortgage paid by all home owners in a town is $2365 with a standard deviation
of $340.
a. Using Chebyshev's theorem, find at least what percentage of all home owners in this town pay a
monthly mortgage of
i. $1685 to $3045 ii. $1345 to $3385
*b. Using Chebyshev's theorem, find the interval that contains the monthly mortgage payments of at
least 84% of all home owners.
3.82 The mean life of a certain brand of auto batteries is 44 months with a standard deviation of 3 months.
Assume that the lives of all auto batteries of this brand have a bell-shaped distribution. Using the empir-
ical rule, find the percentage of auto batteries of this brand that have a life of
a. 41 to 47 months b. 38 to 50 months c. 35 to 53 months
3.83 According to an article in the Washington Post (Washington Post, January 5, 2009), the average em-
ployee share of health insurance premiums at large U.S. companies is expected to be $3423 in 2009. Sup-
pose that the current annual payments by all such employees toward health insurance premiums have a
bell-shaped distribution with a mean of $3423 and a standard deviation of $520. Using the empirical rule,
find the approximate percentage of employees whose annual payments toward such premiums are between
a. $1863 and $4983 b. $2903 and $3943 c. $2383 and $4463
3.84 The prices of all college textbooks follow a bell-shaped distribution with a mean of $105 and a stan-
dard deviation of $20.
a. Using the empirical rule, find the percentage of all college textbooks with their prices between
i. $85 and $125 ii. $65 and $145
*b. Using the empirical rule, find the interval that contains the prices of 99.7% of college textbooks.
3.85 Suppose that on a certain section of 1-95 with a posted speed limit of 65 mph, the speeds of all
vehicles have a bell-shaped distribution with a mean of 72 mph and a standard deviation of 3 mph.
a. Using the empirical rule, find the percentage of vehicles with the following speeds on this sec-
tion of 1-95.
i. 63 to 81 mph ii. 69 to 75 mph
*b. Using the empirical rule, find the interval that contains the speeds of 95% of vehicles traveling
on this section of 1-95.
3.5 Measures of Position
A measure of position determines the position of a single value in relation to other values in
a sample or a population data set. There are many measures of position; however, only quar-
tiles, percentiles, and percentile rank are discussed in this section.
3.5.1 Quartiles and Interquartile Range
Quartiles are the summary measures that divide a ranked data set into four equal parts. Three
measures will divide any data set into four equal parts. These three measures are the first quar-
tile (denoted by giX the second quartile (denoted by Q 2 ), and the third quartile (denoted by
Q 3 ). The data should be ranked in increasing order before the quartiles are determined. The
quartiles are defined as follows.
Definition
Quartiles Quartiles are three summary measures that divide a ranked data set into four equal
parts. The second quartile is the same as the median of a data set. The first quartile is the value
of the middle term among the observations that are less than the median, and the third quartile
is the value of the middle term among the observations that are greater than the median.
Figure 3.11 describes the positions of the three quartiles.
3.5 Measures of Position
111
Each of these portions contains 25% of the Figure 3.1 1 Quartiles.
observations of a data set arranged in increasing order
I I I 4
25%
25%
25%
25%
Q^ Q 2 Qa
Approximately 25% of the values in a ranked data set are less than (2i an d about 75% are
greater than Q x . The second quartile, Q 2 , divides a ranked data set into two equal parts; hence,
the second quartile and the median are the same. Approximately 75% of the data values are less
than Q 3 and about 25% are greater than Q 3 .
The difference between the third quartile and the first quartile for a data set is called the
interquartile range (IQR).
Calculating Interquartile Range The difference between the third and the first quartiles gives the
interquartile range; that is,
IQR = Interquartile range = Q 3 — <2i
Examples 3-20 and 3-21 show the calculation of the quartiles and the interquartile range.
■ EXAMPLE 3-20
Refer to Table 3.3 in Example 3-5, which gives the 2008 profits (rounded to billions of dol-
lars) of 12 companies selected from all over the world. That table is reproduced below.
Finding quartiles and the
interquartile range.
Company
2008 Profits
(billions of dollars)
Merck & Co
8
IBM
12
Unilever
7
Microsoft
17
Petrobras
14
Exxon Mobil
45
Lukoil
10
AT&T
13
Nestle
17
Vodafone
13
Deutsche Bank
9
China Mobile
11
(a) Find the values of the three quartiles. Where does the 2008 profits of Merck & Co
fall in relation to these quartiles?
(b) Find the interquartile range.
Solution
(a) First we rank the given data in increasing order. Then we calculate the three quartiles as
follows:
Values less than the median Values greater than the median
Finding quartiles for an even
number of data values.
I~7 8 9 10 11 12~l | HI 13 14 17 17 4sl
9+10 12+13 14 + 17
g, = —r— = 9.5 Q 2 = = 12.5 Q 3 = = 15.5
2 2 | 2
Also the median
112 Chapter 3 Numerical Descriptive Measures
Finding the interquartile range.
The value of Q 2 , which is also the median, is given by the value of the middle term
in the ranked data set. For the data of this example, this value is the average of the
sixth and seventh terms. Consequently, Q 2 is $12.5 billion. The value of Q l is given
by the value of the middle term of the six values that fall below the median (or Q 2 ).
Thus, it is obtained by taking the average of the third and fourth terms. So, Q x is $9.5
billion. The value of Q 3 is given by the value of the middle term of the six values that
fall above the median. For the data of this example, Q 3 is obtained by taking the av-
erage of the ninth and tenth terms, and it is $15.5 billion.
The value of Q x = $9.5 billion indicates that 25% of the companies in this sample
had 2008 profits less than $9.5 billion and 75% of the companies had 2008 profits
higher than $9.5 billion. Similarly, we can state that half of these companies had 2008
profits less than $12.5 billion and the other half had profits greater than $12.5 billion
since the second quartile is $12.5 billion. The value of Q 3 = $15.5 billion indicates
that 75% of the companies had 2008 profits less than $15.5 billion and 25% had prof-
its greater than this value.
By looking at the position of $8 billion, which is the 2008 profit of Merck & Co,
we can state that this value lies in the bottom 25% of the profits for 2008.
(b) The interquartile range is given by the difference between the values of the third and
the first quartiles. Thus,
IQR = Interquartile range = Q 3 - Q l = 15.5 - 9.5 = $6 billion ■
Finding quartiles and the
interquartile range.
■ EXAMPLE 3-21
The following are the ages (in years) of nine employees of an insurance company:
47 28 39 51 33 37 59 24 33
(a) Find the values of the three quartiles. Where does the age of 28 years fall in relation
to the ages of these employees?
(b) Find the interquartile range.
Finding quartiles for an odd
number of data values.
Solution
(a) First we rank the given data in increasing order. Then we calculate the three quartiles
as follows:
Values less than the median
24
28
33
33
Gi
28 + 33
2
30.5
Values greater than the median
37
t
t
Also the median
39
47
51
59
Q 3
47 + 51
49
Finding the interquartile range.
Thus the values of the three quartiles are
Qi = 30.5 years, Q 2 = 37 years, and g 3 = 49 years
The age of 28 falls in the lowest 25% of the ages,
(b) The interquartile range is
IQR = Interquartile range = g 3 - Q x = 49 - 30.5 = 18.5 years
3.5 Measures of Position
3.5.2 Percentiles and Percentile Rank
Percentiles are the summary measures that divide a ranked data set into 100 equal parts. Each
(ranked) data set has 99 percentiles that divide it into 100 equal parts. The data should be ranked
in increasing order to compute percentiles. The kth percentile is denoted by P k , where k is an
integer in the range 1 to 99. For instance, the 25th percentile is denoted by P 25 - Figure 3.12
shows the positions of the 99 percentiles.
Each of these portions contains 1% of the observations of a data
ill set arranged in increasing order | | |
. T , T . T . 3J , T . T . T
1%
1%
1%
I
1% 1% 1%
p -\ P 2 P 3 P 97 P 9S P 99
Figure 3.12 Percentiles.
Thus, the A:th percentile, P h can be defined as a value in a data set such that about k% of
the measurements are smaller than the value of P k and about (100 — k)% of the measurements
are greater than the value of P k .
The approximate value of the kth percentile is determined as explained next.
Calculating Percentiles The (approximate) value of the kth percentile, denoted by P k , is
kn
P t = Value of the th term in a ranked data set
.100,
where k denotes the number of the percentile and n represents the sample size.
Example 3-22 describes the procedure to calculate the percentiles.
■ EXAMPLE 3-22
Refer to the data on 2008 profits for 12 companies given in Example 3-20. Find the value of
Fiiiuins the percentile for o
the 42nd percentile. Give a brief interpretation of the 42nd percentile. rfafa se(
Solution From Example 3-20, the data arranged in increasing order are as follows:
7 8 9 10 11 12 13 13 14 17 17 45
The position of the 42nd percentile is
kn 42(12)
100 100
5.04th term
The value of the 5.04th term can be approximated by the value of the 5th term in the ranked
data. Therefore,
P 42 = 42nd percentile = 11 = $11 billion
Thus, approximately 42% of these 12 companies had 2008 profits less than or equal to
$11 billion. ■
We can also calculate the percentile rank for a particular value x, of a data set by using
the formula given below. The percentile rank of x, gives the percentage of values in the data set
that are less than x,-.
Finding Percentile Rank of a Value
Number of values less than x,
Percentile rank of x, =
X 100
Total number of values in the data set
Example 3-23 shows how the percentile rank is calculated for a data value.
114 Chapter 3 Numerical Descriptive Measures
Finding the percentile rank ^ EXAMPLE 3 23
for a data value. Refer to the data on 2008 profits for 12 companies given in Example 3-20. Find the percentile
rank for $14 billion profit of Petrobras. Give a brief interpretation of this percentile rank.
Solution From Example 3-20, the data arranged in increasing order are as follows:
7 8 9 10 11 12 13 13 14 17 17 45
In this data set, 8 of the 12 values are less than $14 billion. Hence,
8
Percentile rank of 14 = — X 100 = 66.67%
12
Rounding this answer to the nearest integral value, we can state that about 67% of the com-
panies in these 12 had less than $14 billion profits in 2008. Hence, 33% of these 12 compa-
nies had $14 billion or higher profits in 2008. I
EXERCISES
CONCEPTS AND PROCEDURES
3.86 Briefly describe how the three quartiles are calculated for a data set. Illustrate by calculating the three
quartiles for two examples, the first with an odd number of observations and the second with an even num-
ber of observations.
3.87 Explain how the interquartile range is calculated. Give one example.
3.88 Briefly describe how the percentiles are calculated for a data set.
3.89 Explain the concept of the percentile rank for an observation of a data set.
■ APPLICATIONS
3.90 The following data give the weights (in pounds) lost by 15 members of a health club at the end of
2 months after joining the club.
5 10 8 7 25 12 5 14
11 10 21 9 8 11 18
a. Compute the values of the three quartiles and the interquartile range.
b. Calculate the (approximate) value of the 82nd percentile.
c. Find the percentile rank of 10.
3.91 The following data give the speeds of 13 cars (in mph) measured by radar, traveling on 1-84.
73 75 69 68 78 69 74
76 72 79 68 77 71
a. Find the values of the three quartiles and the interquartile range.
b. Calculate the (approximate) value of the 35th percentile.
c. Compute the percentile rank of 7 1 .
3.92 The following data give the numbers of computer keyboards assembled at the Twentieth Century
Electronics Company for a sample of 25 days.
45
52
48
41
56
46
44
42
48
53
51
53
51
48
46
43
52
50
54
47
44
47
50
49
52
a. Calculate the values of the three quartiles and the interquartile range.
b. Determine the (approximate) value of the 53rd percentile.
c. Find the percentile rank of 50.
3.93 The following data give the numbers of minor penalties accrued by each of the 30 National Hockey
League franchises during the 2007-08 regular season.
318
336
337
339
362
363
366
369
372
375
378
381
384
385
386
387
390
393
395
403
405
409
417
431
433
434
438
444
461
480
3.6 Box-and-Whisker Plot 115
a. Calculate the values of the three quartiles and the interquartile range.
b. Find the approximate value of the 57th percentile.
c. Calculate the percentile rank of 417.
3.94 Refer to Exercise 3.22. The following data give the number of students suspended for bringing
weapons to schools in the Tri-City School District for each of the past 12 weeks.
1 59 1 21 17 6 9 1 01 43 6 5
a. Determine the values of the three quartiles and the interquartile range. Where does the value of
10 fall in relation to these quartiles?
b. Calculate the (approximate) value of the 55th percentile.
c. Find the percentile rank of 7.
3.95 Nixon Corporation manufactures computer monitors. The following data give the numbers of com-
puter monitors produced at the company for a sample of 30 days.
24
32
27
23
33
33
29
25
23
36
26
26
31
20
27
33
27
23
28
29
31
35
34
22
37
28
23
35
31
43
a. Calculate the values of the three quartiles and the interquartile range. Where does the value of
31 lie in relation to these quartiles?
b. Find the (approximate) value of the 65th percentile. Give a brief interpretation of this percentile.
c. For what percentage of the days was the number of computer monitors produced 32 or higher?
Answer by finding the percentile rank of 32.
3.96 The following data give the numbers of new cars sold at a dealership during a 20-day period.
8 5 12 3 9 10 6 12 8 8
4 16 10 11 7 7 3 5 9 11
a. Calculate the values of the three quartiles and the interquartile range. Where does the value of 4
lie in relation to these quartiles?
b. Find the (approximate) value of the 25th percentile. Give a brief interpretation of this percentile.
c. Find the percentile rank of 10. Give a brief interpretation of this percentile rank.
3.97 According to Fair Isaac, "The Median FICO (Credit) Score in the U.S. is 723" (The Credit Scoring
Site, 2009). Suppose the following data represent the credit scores of 22 randomly selected loan applicants.
494 728 468 533 747 639 430 690 604 422 356
805 749 600 797 702 628 625 617 647 772 572
a. Calculate the values of the three quartiles and the interquartile range. Where does the value 617
fall in relation to these quartiles?
b. Find the approximate value of the 30th percentile. Give a brief interpretation of this percentile.
c. Calculate the percentile rank of 533. Give a brief interpretation of this percentile rank.
3.6 Box-and-Whisker Plot
A box-and-whisker plot gives a graphic presentation of data using five measures: the median,
the first quartile, the third quartile, and the smallest and the largest values in the data set be-
tween the lower and the upper inner fences. (The inner fences are explained in Example 3-24
below.) A box-and-whisker plot can help us visualize the center, the spread, and the skewness
of a data set. It also helps detect outliers. We can compare different distributions by making
box-and-whisker plots for each of them.
Definition
Box-and-Whisker Plot A plot that shows the center, spread, and skewness of a data set. It is con-
structed by drawing a box and two whiskers that use the median, the first quartile, the third quar-
tile, and the smallest and the largest values in the data set between the lower and the upper inner
fences.
Example 3-24 explains all the steps needed to make a box-and-whisker plot.
Chapter 3 Numerical Descriptive Measures
Constructing a
box-and-whisker plot.
■ EXAMPLE 3-24
The following data are the incomes (in thousands of dollars) for a sample of 12 house-
holds.
75 69 84 112 74 104 81 90 94 144 79 98
Construct a box-and-whisker plot for these data.
Solution The following five steps are performed to construct a box-and-whisker plot.
Step 1. First, rank the data in increasing order and calculate the values of the median, the
first quartile, the third quartile, and the interquartile range. The ranked data are
69 74 75 79 81 84 90 94 98 104 112 144
For these data,
Median = (84 + 90)/2 = 87
Qi = (75 + 79)/2 = 77
Q 3 = (98 + 104)/2 = 101
IQR = <3 3 - Qi = 101 - 77 = 24
Step 2. Find the points that are 1.5 X IQR below g, and 1.5 X IQR above Q 3 . These
two points are called the lower and the upper inner fences, respectively.
1.5 X IQR = 1.5 X 24 = 36
Lower inner fence = Q x — 36 = 77 — 36 = 41
Upper inner fence = Q 3 + 36 = 101 + 36 = 137
Step 3. Determine the smallest and the largest values in the given data set within the two
inner fences. These two values for our example are as follows:
Smallest value within the two inner fences = 69
Largest value within the two inner fences = 112
Step 4. Draw a horizontal line and mark the income levels on it such that all the values
in the given data set are covered. Above the horizontal line, draw a box with its left side
at the position of the first quartile and the right side at the position of the third quartile. In-
side the box, draw a vertical line at the position of the median. The result of this step is
shown in Figure 3.13.
L
First Third
quartile quartile
Median
J
65 75 85 95 105 115 125 135 145
Income
Figure 3.13
Step 5. By drawing two lines, join the points of the smallest and the largest values within
the two inner fences to the box. These values are 69 and 1 12 in this example as listed in
Step 3. The two lines that join the box to these two values are called whiskers. A value
that falls outside the two inner fences is shown by marking an asterisk and is called an out-
lier. This completes the box-and-whisker plot, as shown in Figure 3.14.
3.6 Box-and-Whisker Plot 117
First
quartile
Smallest value
within the two
inner fences
65
Third
quartile
1
Median
r— Large
within
T inner
Largest value
the two
fences
An
outlier
75
85
95 105 115
Income
J
125 135 145
Figure 3.14
In Figure 3.14, about 50% of the data values fall within the box, about 25% of the val-
ues fall on the left side of the box, and about 25% fall on the right side of the box. Also, 50%
of the values fall on the left side of the median and 50% lie on the right side of the median.
The data of this example are skewed to the right because the lower 50% of the values are
spread over a smaller range than the upper 50% of the values. H
The observations that fall outside the two inner fences are called outliers. These outliers
can be classified into two kinds of outliers — mild and extreme outliers. To do so, we define two
outer fences — a lower outer fence at 3.0 X IQR below the first quartile and an upper outer
fence at 3.0 X IQR above the third quartile. If an observation is outside either of the two inner
fences but within either of the two outer fences, it is called a mild outlier. An observation that
is outside either of the two outer fences is called an extreme outlier. For the previous example,
the outer fences are at 5 and 173. Because 144 is outside the upper inner fence but inside the
upper outer fence, it is a mild outlier.
For a symmetric data set, the line representing the median will be in the middle of the box
and the spread of the values will be over almost the same range on both sides of the box.
EXERCISES
CONCEPTS AND PROCEDURES
3.98 Briefly explain what summary measures are used to construct a box-and-whisker plot.
3.99 Prepare a box-and-whisker plot for the following data:
36
43
28
52
41
59
47
61
24
55
63
73
32
25
35
49
31
22
61
42
58
65
98
34
Does this data set contain any outliers?
3.100 Prepare a box-and-whisker plot for the following data:
11 8 26 31 62 19 7
33 30 42 15 18 23 29
Does this data set contain any outliers?
3
13
14
16
75
6
■ APPLICATIONS
3.101 The following data give the time (in minutes) that each of 20 students selected from a university
waited in line at their bookstore to pay for their textbooks in the beginning of the Fall 2009 semester.
15 8 23 21 5 17 31 22 34 6
5 10 14 17 16 25 30 3 31 19
Prepare a box-and-whisker plot. Comment on the skewness of these data.
3.102 Refer to Exercise 3.97. The following data represent the credit scores of 22 randomly selected loan
applicants.
494 728 468 533 747 639 430 690 604 422 356
805 749 600 797 702 628 625 617 647 772 572
Prepare a box-and-whisker plot. Are these data skewed in any direction?
118 Chapter 3 Numerical Descriptive Measures
3.103 The following data give the recent estimates of crude oil reserves (in billions of barrels) of Saudi
Arabia, Iraq, Kuwait, Iran, United Arab Emirates, Venezuela, Russia, Libya, Nigeria, China, Mexico, and
the United States. The reserves for these countries are listed in that order.
261.7 112.0 97.7 94.4 80.3 64.0
51.2 29.8 27.0 26.8 25.0 22.5
Prepare a box-and-whisker plot. Are the data symmetric or skewed?
3.104 The following data give the numbers of computer keyboards assembled at the Twentieth Century
Electronics Company for a sample of 25 days.
45
52
48
41
56
46
44
42
48
53
51
53
51
48
46
43
52
50
54
47
44
47
50
49
52
Prepare a box-and-whisker plot. Comment on the skewness of these data.
3.105 Refer to Exercise 3.93. The following data represent the numbers of minor penalties accrued by
each of the 30 National Hockey League franchises during the 2007-08 regular season.
318
336
337
339
362
363
366
369
372
375
378
381
384
385
386
387
390
393
395
403
405
409
417
431
433
434
438
444
461
480
Prepare a box-and-whisker plot. Are these data skewed in any direction?
3.106 Refer to Exercise 3.22. The following data give the number of students suspended for bringing
weapons to schools in the Tri-City School District for each of the past 12 weeks.
1 59 1 21 17 6 9 1 01 43 6 5
Make a box-and-whisker plot. Comment on the skewness of these data.
3.107 Nixon Corporation manufactures computer monitors. The following are the numbers of computer
monitors produced at the company for a sample of 30 days:
24
32
27
23
33
33
29
25
23
28
21
26
31
20
27
33
27
23
28
29
31
35
34
22
26
28
23
35
31
27
Prepare a box-and-whisker plot. Comment on the skewness of these data.
3.108 The following data give the numbers of new cars sold at a dealership during a 20-day period.
8 5 1 2 3 9 1 06 1 28 8
4 1 61 01 17 7 3 5 9 1 1
Make a box-and-whisker plot. Comment on the skewness of these data.
USES AND MISUSES... UNEMPLOYMENT RATES
During his fiscal year 2010 budget address, New Jersey Governor Jon
Corzine made the following statement: "Our unemployment rate is
below the national average. And as of January, at least 18 states had
higher jobless rates than New Jersey— often, significantly higher" (The
2010 State of New Jersey Budget Address, March 10, 2009).
Governor Corzine is certainly not the only politician to reference
statistics during a campaign speech, budget address, or State of the State
address. As a statistically literate member of society, one should be able
to examine a statement and the data to determine its meaning.
According to the Bureau of Labor Statistics, the unemployment
rate in New Jersey was 7.3% in January 2009. The rate was ranked
28th out of the 50 states and the District of Columbia, which was
ranked 43rd with an unemployment rate of 9.2%. Hence, 22 states
had higher jobless rates than New Jersey (BLS, January, 2009).
Governor Corzine stated that the state unemployment rate is be-
low the national average. Based on what you have learned in this
chapter, you might try to obtain the national average by taking the av-
erage of the 51 unemployment rates. However, the average of these
51 numbers was 7.198 as of January 2009, which is not greater than
7.3. One must wonder why the Governor's statement is seemingly false.
Each state's unemployment rate is calculated by finding the
percentage of the state's labor force that is currently unable to find
a job. States such as California, New York, and Texas have many
more people in their labor forces than states such as Wyoming,
North Dakota, and Rhode Island; however, unemployment rates are
on the same scale, primarily for purposes of comparison. Whenever
an average is calculated, each observation contributes equally to the
average, which means that Wyoming and California would con-
tribute equally to the national unemployment rate, when California
should have a much larger impact on the national rate.
When the national unemployment rate is calculated, it meas-
ures the percentage of the country's labor force that is currently un-
Glossary 119
able to obtain a job. The states with larger labor forces will have a
greater impact on the national rate than states with smaller labor
forces. The national unemployment rate in January 2009 was 7.6%,
which is higher than New Jersey's 7.3% unemployment rate. Hence,
the statement made by Governor Corzine (or his speechwriters) is
correct in its intended meaning. However, the statistical terminology
used in the speech is inappropriate.
The other piece of statistical terminology used by Governor
Corzine is associated with unemployment rates being "significantly
higher." The unemployment rates that were higher than 7.3% ranged
from 7.4% (Massachusetts) to 11.6% (Michigan). How much higher
than 7.3% must an unemployment rate be in order to be classified
as "significantly higher?" Would someone who is unemployed answer
this question differently than someone who has a job? In Chapter 9,
we will address the notion of statistical significance, which will pro-
vide us with a method for answering this question.
One other question that is often raised when statistics are
used in a speech or analysis relates to whether the comparisons
made in the analysis are appropriate. For example. New Jersey has
a large number of pharmaceutical firms. Southern New Jersey has
a large agricultural sector. Many statisticians would argue that it
makes more sense to compare agricultural unemployment levels
across states, as well as pharmaceutical (or science and technol-
ogy) unemployment rates, instead of combining them. Choosing
a level of aggregation can have dramatic impact on the interpre-
tation of data.
Glossary
Bimodal distribution A distribution that has two modes.
Box-and-whisker plot A plot that shows the center, spread, and
skewness of a data set with a box and two whiskers using the median,
the first quartile, the third quartile, and the smallest and the largest
values in the data set between the lower and the upper inner fences.
Chebyshev's theorem For any number k greater than 1, at least
(1 — 1/k 2 ) of the values for any distribution lie within k standard de-
viations of the mean.
Coefficient of variation A measure of relative variability that ex-
presses standard deviation as a percentage of the mean.
Empirical rule For a specific bell-shaped distribution, about 68%
of the observations fall in the interval (ju — a) to (/jl + a), about
95% fall in the interval Ou, - 2a) to (/x + 2a), and about 99.7% fall
in the interval (/x — 3a) to (ii + 3a).
First quartile The value in a ranked data set such that about 25%
of the measurements are smaller than this value and about 75% are
larger. It is the median of the values that are smaller than the me-
dian of the whole data set.
Geometric mean Calculated by taking the nth root of the product
of all values in a data set.
Interquartile range (IQR) The difference between the third and
the first quartiles.
Lower inner fence The value in a data set that is 1.5 X IQR be-
low the first quartile.
Lower outer fence The value in a data set that is 3.0 X IQR be-
low the first quartile.
Mean A measure of central tendency calculated by dividing the
sum of all values by the number of values in the data set.
Measures of central tendency Measures that describe the center
of a distribution. The mean, median, and mode are three of the meas-
ures of central tendency.
Measures of dispersion Measures that give the spread of a distri-
bution. The range, variance, and standard deviation are three such
measures.
Measures of position Measures that determine the position of a
single value in relation to other values in a data set. Quartiles, per-
centiles, and percentile rank are examples of measures of position.
Median The value of the middle term in a ranked data set. The
median divides a ranked data set into two equal parts.
Mode The value (or values) that occurs with highest frequency in
a data set.
Multimodal distribution A distribution that has more than two
modes.
Parameter A summary measure calculated for population data.
Percentile rank The percentile rank of a value gives the percent-
age of values in the data set that are smaller than this value.
Percentiles Ninety-nine values that divide a ranked data set into
100 equal parts.
Quartiles Three summary measures that divide a ranked data set
into four equal parts.
Range A measure of spread obtained by taking the difference be-
tween the largest and the smallest values in a data set.
Second quartile Middle or second of the three quartiles that divide
a ranked data set into four equal parts. About 50% of the values in
the data set are smaller and about 50% are larger than the second
quartile. The second quartile is the same as the median.
Standard deviation A measure of spread that is given by the pos-
itive square root of the variance.
Statistic A summary measure calculated for sample data.
Third quartile Third of the three quartiles that divide a ranked
data set into four equal parts. About 75% of the values in a data set
are smaller than the value of the third quartile and about 25% are
larger. It is the median of the values that are greater than the median
of the whole data set.
Trimmed mean The k% trimmed mean is obtained by dropping
k% of the smallest values and k% of the largest values from the given
data and then calculating the mean of the remaining (100 — 2k)%
of the values.
120 Chapter 3 Numerical Descriptive Measures
Unimodal distribution A distribution that has only one mode.
Upper inner fence The value in a data set that is 1.5 X IQR above
the third quartile.
Upper outer fence The value in a data set that is 3.0 X IQR above
the third quartile.
Variance A measure of spread.
Weighted mean Mean of a data set whose values are assigned dif-
ferent weights before the mean is calculated.
Supplementary Exercises
3.109 Each year the faculty at Metro Business College chooses 10 members from the current graduating
class that they feel are most likely to succeed. The data below give the current annual incomes (in thou-
sands of dollars) of the 10 members of the class of 2000 who were voted most likely to succeed.
59 68 84 78 107 382 56 74 97 60
a. Calculate the mean and median.
b. Does this data set contain any outlier(s)? If yes, drop the outlier(s) and recalculate the mean and
median. Which of these measures changes by a greater amount when you drop the outlier(s)?
c. Is the mean or the median a better summary measure for these data? Explain.
3.110 The Belmont Stakes is the final race in the annual Triple Crown of thoroughbred racing. The race
is 1.5 miles in length, and the record for the fastest time (2:24.00 seconds) was set by Secretariat in 1973.
The following data come from the 1999 to 2008 Belmont Stakes winners and show how much faster Sec-
retariat's record time is compared to these winning times (in seconds) for the years 1999 to 2008.
3.80 7.20 2.80 5.71 4.26 3.50 4.75 3.81 4.74 5.65
a. Calculate the mean and median. Do these data have a mode? Why or why not?
b. Compute the variance, standard deviation, and range for these data.
3.111 The following table gives the total points scored by each of the top 10 scorers during the 2007-08
regular season of National Basketball Association (Source: www.NBA.com.)
Player
Points Scored
Kobe Bryant
2323
LeBron James
2250
Allen Iverson
2164
Amare Stoudemire
1989
Carmelo Anthony
1978
Richard Jefferson
1857
Dirk Nowitski
1817
Baron Davis
1791
Jason Richardson
1788
Joe Johnson
1779
a. Calculate the mean and median. Do these data have a mode? Why or why not?
b. Compute the variance, standard deviation, and range for these data.
3.112 The following data give the numbers of driving citations received by 12 drivers.
4 8 3 1 17 4 1 48 1 37 9
a. Find the mean, median, and mode for these data.
b. Calculate the range, variance, and standard deviation.
c. Are the values of the summary measures in parts a and b population parameters or sample
statistics?
3.113 The following table gives the distribution of the amounts of rainfall (in inches) for July 2009 for
50 cities.
Supplementary Exercises 121
R'tinf'ill
IVtl 1111 till
ill LI1I1 1IC1 1)1 v.- l Lies
to less than 2
6
2 to less than 4
1 A
4 to less than 6
20
6 to less than 8
7
8 to less than 10
4
10 to less than 12
3
Find the mean, variance, and standard deviation. Are the values of these summary measures population
parameters or sample statistics?
3.114 The following table gives the frequency distribution of the times (in minutes) that 50 commuter students
at a large university spent looking for parking spaces on the first day of classes in the Fall semester of 2009.
Time
Number of Students
to less than 4
1
4 to less than 8
7
8 to less than 12
15
12 to less than 16
18
16 to less than 20
6
20 to less than 24
3
Find the mean, variance, and standard deviation. Are the values of these summary measures population
parameters or sample statistics?
3.115 The mean time taken to learn the basics of a word processor by all students is 200 minutes with a
standard deviation of 20 minutes.
a. Using Chebyshev's theorem, find at least what percentage of students will learn the basics of
this word processor in
i. 160 to 240 minutes ii. 140 to 260 minutes
*b. Using Chebyshev's theorem, find the interval that contains the time taken by at least 75% of
all students to learn this word processor.
3.116 According to an estimate, Americans were expected to spend an average of 1669 hours watching
television in 2004. Assume that the average time spent watching television by Americans in 2009 has a
distribution that is skewed to the right with a mean of 1750 hours and a standard deviation of 450 hours.
a. Using Chebyshev's theorem, find at least what percentage of Americans watched television in
2009 for
i. 850 to 2650 hours ii. 400 to 3100 hours
*b. Using Chebyshev's theorem, find the interval that will contain the television viewing times of
at least 84% of all Americans.
3.117 Refer to Exercise 3.1 15. Suppose the times taken to learn the basics of this word processor by all stu-
dents have a bell-shaped distribution with a mean of 200 minutes and a standard deviation of 20 minutes.
a. Using the empirical rule, find the percentage of students who will learn the basics of this
word processor in
i. 180 to 220 minutes ii. 160 to 240 minutes
*b. Using the empirical rule, find the interval that contains the time taken by 99.7% of all stu-
dents to learn this word processor.
3.118 Assume that the annual earnings of all employees with CPA certification and 6 years of experience
and working for large firms have a bell-shaped distribution with a mean of $134,000 and a standard de-
viation of $12,000.
a. Using the empirical rule, find the percentage of all such employees whose annual earnings are
between
i. $98,000 and $170,000 ii. $110,000 and $158,000
*b. Using the empirical rule, find the interval that contains the annual earnings of 68% of all such
employees.
122 Chapter 3 Numerical Descriptive Measures
3.119 Refer to the data of Exercise 3.109 on the current annual incomes (in thousands of dollars) of the
10 members of the class of 2000 of the Metro Business College who were voted most likely to succeed.
59 68 84 78 107 382 56 74 97 60
a. Determine the values of the three quartiles and the interquartile range. Where does the value of
74 fall in relation to these quartiles?
b. Calculate the (approximate) value of the 70th percentile. Give a brief interpretation of this per-
centile.
c. Find the percentile rank of 97. Give a brief interpretation of this percentile rank.
3.120 Refer to the data given in Exercise 3.111 on the total points scored by the top 10 NBA scorers dur-
ing the 2007-08 NBA regular season.
a. Calculate the values of the three quartiles and the interquartile range. Where does the value
1978 fall in relation to these quartiles?
b. Find the approximate value of the 70th percentile. Give a brief interpretation of this percentile.
c. Calculate the percentile rank of 1788. Give a brief interpretation of this percentile rank.
3.121 A student washes her clothes at a laundromat once a week. The data below give the time (in min-
utes) she spent in the laundromat for each of 15 randomly selected weeks. Here, time spent in the laun-
dromat includes the time spent waiting for a machine to become available.
75 62 84 73 107 81 93 72
135 77 85 67 90 83 112
Prepare a box-and-whisker plot. Is the data set skewed in any direction? If yes, is it skewed to the right
or to the left? Does this data set contain any outliers?
3.122 The following data give the lengths of time (in weeks) taken to find a full-time job by 18 computer
science majors who graduated in 2008 from a small college.
30 43 32 21 65 8 4 18 16
38 9 44 33 23 24 81 42 55
Make a box-and-whisker plot. Comment on the skewness of this data set. Does this data set contain any
outliers?
Advanced Exercises
3.123 Melissa's grade in her math class is determined by three 100-point tests and a 200-point final exam.
To determine the grade for a student in this class, the instructor will add the four scores together and di-
vide this sum by 5 to obtain a percentage. This percentage must be at least 80 for a grade of B. If Melissa's
three test scores are 75, 69, and 87, what is the minimum score she needs on the final exam to obtain a
B grade?
3.124 Jeffrey is serving on a six-person jury for a personal-injury lawsuit. All six jurors want to award
damages to the plaintiff but cannot agree on the amount of the award. The jurors have decided that each
of them will suggest an amount that he or she thinks should be awarded; then they will use the mean of
these six numbers as the award to recommend to the plaintiff.
a. Jeffrey thinks the plaintiff should receive $20,000, but he thinks the mean of the other five
jurors' recommendations will be about $12,000. He decides to suggest an inflated amount so
that the mean for all six jurors is $20,000. What amount would Jeffrey have to suggest?
b. How might this jury revise its procedure to prevent a juror like Jeffrey from having an undue
influence on the amount of damages to be awarded to the plaintiff?
3.125 The heights of five starting players on a basketball team have a mean of 76 inches, a median of
78 inches, and a range of 11 inches.
a. If the tallest of these five players is replaced by a substitute who is 2 inches taller, find the new
mean, median, and range.
b. If the tallest player is replaced by a substitute who is 4 inches shorter, which of the new values
(mean, median, range) could you determine, and what would their new values be?
3.126 On a 300-mile auto trip, Lisa averaged 52 mph for the first 100 miles, 65 mph for the second 100
miles, and 58 mph for the last 100 miles.
a. How long did the 300-mile trip take?
b. Could you find Lisa's average speed for the 300-mile trip by calculating (52 + 65 + 58)/3? If
not, find the correct average speed for the trip.
Supplementary Exercises 123
3.127 A small country bought oil from three different sources in one week, as shown in the following table.
Source
Barrels Purchased
Price per Barrel ($)
Mexico
1000
51
Kuwait
200
64
Spot Market
100
70
Find the mean price per barrel for all 1 300 barrels of oil purchased in that week.
3.128 During the 2008 winter season, a homeowner received four deliveries of heating oil, as shown in
the following table.
Gallons Purchased
Price per Gallon ($)
209
2.22
182
2.34
157
2.41
149
2.43
The homeowner claimed that the mean price he paid for oil during the season was (2.22 + 2.34 + 2.41 +
2.43)/4 = $2.35 per gallon. Do you agree with this claim? If not, explain why this method of calculating
the mean is not appropriate in this case and find the correct value of the mean price.
3.129 In the Olympic Games, when events require a subjective judgment of an athlete's performance, the
highest and lowest of the judges' scores may be dropped. Consider a gymnast whose performance is judged
by seven judges and the highest and the lowest of the seven scores are dropped.
a. Gymnast A's scores in this event are 9.4, 9.7, 9.5, 9.5, 9.4, 9.6, and 9.5. Find this gymnast's
mean score after dropping the highest and the lowest scores.
b. The answer to part a is an example of (approximately) what percentage of trimmed mean?
c. Write another set of scores for a gymnast B so that gymnast A has a higher mean score than
gymnast B based on the trimmed mean, but gymnast B would win if all seven scores were
counted. Do not use any scores lower than 9.0.
3.130 A survey of young people's shopping habits in a small city during the summer months of 2009
showed the following: Shoppers aged 12 to 14 years took an average of 8 shopping trips per month and
spent an average of $14 per trip. Shoppers aged 15 to 17 years took an average of 11 trips per month and
spent an average of $18 per trip. Assume that this city has 1 100 shoppers aged 12 to 14 years and 900 shop-
pers aged 15 to 17 years.
a. Find the total amount spent per month by all these 2000 shoppers in both age groups.
b. Find the mean number of shopping trips per person per month for these 2000 shoppers.
c. Find the mean amount spent per person per month by shoppers aged 12 to 17 years in this city.
3.131 The following table shows the total population and the number of deaths (in thousands) due to heart
attack for two age groups (in years) in Countries A and B for 2009.
Age 30 and Under
Age 31 and Over
A
B
A
B
Population
40,000
25,000
20,000
35,000
Deaths due to heart attack
1000
500
2000
3000
a. Calculate the death rate due to heart attack per 1000 population for the 30 years and under age
group for each of the two countries. Which country has the lower death rate in this age group?
b. Calculate the death rates due to heart attack for the two countries for the 3 1 years and over age
group. Which country has the lower death rate in this age group?
c. Calculate the death rate due to heart attack for the entire population of Country A; then do the
same for Country B. Which country has the lower overall death rate?
d. How can the country with lower death rate in both age groups have the higher overall death
rate? (This phenomenon is known as Simpson's paradox.)
124 Chapter 3 Numerical Descriptive Measures
3.132 In a study of distances traveled to a college by commuting students, data from 100 commuters
yielded a mean of 8.73 miles. After the mean was calculated, data came in late from three students, with
respective distances of 11.5, 7.6, and 10.0 miles. Calculate the mean distance for all 103 students.
3.133 The test scores for a large statistics class have an unknown distribution with a mean of 70 and a
standard deviation of 10.
a. Find k so that at least 50% of the scores are within k standard deviations of the mean.
b. Find k so that at most 10% of the scores are more than k standard deviations above the mean.
3.134 The test scores for a very large statistics class have a bell-shaped distribution with a mean of 70 points.
a. If 16% of all students in the class scored above 85, what is the standard deviation of the scores?
b. If 95% of the scores are between 60 and 80, what is the standard deviation?
3.135 How much does the typical American family spend to go away on vacation each year? Twenty-five
randomly selected households reported the following vacation expenditures (rounded to the nearest hun-
dred dollars) during the past year:
2500
500
800
100
200
2200
200
1000
900
321,500
400
500
100
8200
900
1700
1100
600
3400
a. Using both graphical and numerical methods, organize and interpret these data.
b. What measure of central tendency best answers the original question?
3.136 Actuaries at an insurance company must determine a premium for a new type of insurance. A
random sample of 40 potential purchasers of this type of insurance were found to have suffered the
following values of losses (in dollars) during the past year. These losses would have been covered by
the insurance if it were available.
100
32
470
50
14,589
212
93
1127
421
87
135
420
250
12
309
177
295
501
143
167
398
54
141
3709
122
a. Find the mean, median, and mode of these 40 losses.
b. Which of the mean, median, or mode is largest?
c. Draw a box-and-whisker plot for these data, and describe the skewness, if any.
d. Which measure of central tendency should the actuaries use to determine the premium for this
insurance?
3.137 A local golf club has men's and women's summer leagues. The following data give the scores for
a round of 18 holes of golf for 17 men and 15 women randomly selected from their respective leagues.
Men
87
68
92
79
83
67
71
92
112
75
77
102
79
78
85
75
72
Women
101
100
87
95
98
81
117
107
103
97
90
100
99
94
94
a. Make a box-and-whisker plot for each of the data sets and use them to discuss the similarities
and differences between the scores of the men and women golfers.
b. Compute the various descriptive measures you have learned for each sample. How do they
compare?
3.138 Answer the following questions.
a. The total weight of all pieces of luggage loaded onto an airplane is 12,372 pounds, which works
out to be an average of 51.55 pounds per piece. How many pieces of luggage are on the plane?
b. A group of seven friends, having just gotten back a chemistry exam, discuss their scores. Six of
the students reveal that they received grades of 81, 75, 93, 88, 82, and 85, respectively, but the
seventh student is reluctant to say what grade she received. After some calculation she an-
nounces that the group averaged 81 on the exam. What is her score?
Supplementary Exercises 125
3.139 Suppose that there are 150 freshmen engineering majors at a college and each of them will take the
same five courses next semester. Four of these courses will be taught in small sections of 25 students each,
whereas the fifth course will be taught in one section containing all 150 freshmen. To accommodate all
150 students, there must be six sections of each of the four courses taught in 25-student sections. Thus,
there are 24 classes of 25 students each and one class of 150 students.
a. Find the mean size of these 25 classes.
b. Find the mean class size from a student's point of view, noting that each student has five
classes containing 25, 25, 25, 25, and 150 students, respectively.
Are the means in parts a and b equal? If not, why not?
3.140 The following data give the weights (in pounds) of a random sample of 44 college students. (Here
F and M indicate female and male, respectively.)
123 F
195 M
138 M
115 F
179 M
119 F
148 F
147 F
180 M
146 F
179 M
189 M
175 M
108 F
193 M
114 F
179 M
147 M
108 F
128 F
164 F
174 M
128 F
159 M
193 M
204 M
125 F
133 F
115 F
168 M
123 F
183 M
116 F
182 M
174 M
102 F
123 F
99 F
161 M
162 M
155 F
202 M
110 F
132 M
Compute the mean, median, and standard deviation for the weights of all students, of men only, and of
women only. Of the mean and median, which is the more informative measure of central tendency? Write
a brief note comparing the three measures for all students, men only, and women only.
3.141 The distribution of the lengths of fish in a certain lake is not known, but it is definitely not bell
shaped. It is estimated that the mean length is 6 inches with a standard deviation of 2 inches.
a. At least what proportion of fish in the lake are between 3 inches and 9 inches long?
b. What is the smallest interval that will contain the lengths of at least 84% of the fish?
c. Find an interval so that fewer than 36% of the fish have lengths outside this interval.
3.142 The following stem-and-leaf diagram gives the distances (in thousands of miles) driven during the
past year by a sample of drivers in a city.
3 6 9
1 2 8 5 1 5
2 5 16
3 8
4 1
5
6 2
a. Compute the sample mean, median, and mode for the data on distances driven.
b. Compute the range, variance, and standard deviation for these data.
c. Compute the first and third quartiles.
d. Compute the interquartile range. Describe what properties the interquartile range has. When
would it be preferable to using the standard deviation when measuring variation?
3.143 Refer to the data in Problem 3.140. Two individuals, one from Canada and one from England, are
interested in your analysis of these data but they need your results in different units. The Canadian indi-
vidual wants the results in grams (1 pound = 435.59 grams), while the English individual wants the re-
sults in stone (1 stone = 14 pounds).
a. Convert the data on weights from pounds to grams, and then recalculate the mean, median, and
standard deviation of weight for males and females separately. Repeat the procedure, changing
the unit from pounds to stones.
b. Convert your answers from Problem 3.140 to grams and stone. What do you notice about these
answers and your answers from part a?
c. What happens to the values of the mean, median, and standard deviation when you convert
from a larger unit to a smaller unit (e.g., from pounds to grams)? Does the same thing happen
if you convert from a smaller unit (e.g., pounds) to a larger unit (e.g., stone)?
126 Chapter 3 Numerical Descriptive Measures
Figure 3.15 Stacked dotplot of weights in stone and pounds.
d. Figure 3.15 gives a stacked dotplot of these weights in pounds and stone. Which of these two
distributions has more variability? Use your results from parts a to c to explain why this is
the case.
e. Now consider the weights in pounds and grams. Make a stacked dotplot for these data and an-
swer part d.
3.144 Although the standard workweek is 40 hours a week, many people work a lot more than 40 hours
a week. The following data give the numbers of hours worked last week by 50 people.
40.5 41.3
43.7 43.9
47.8 48.2
50.8 51.5
54.4 54.8
41.4 41.5
45.0 45.0
48.3 48.8
51.5 52.3
55.0 55.4
42.0 42.2
45.2 45.8
49.0 49.2
52.3 52.6
55.4 55.4
42.4 42.4
45.9 46.2
49.9 50.1
52.7 52.7
56.2 56.3
42.6 43.3
47.2 47.5
50.6 50.6
53.4 53.9
57.8 58.7
a. The sample mean and sample standard deviation for this data set are 49.012 and 5.080, respec-
tively. Using Chebyshev's theorem, calculate the intervals that contain at least 75%, 88.89%,
and 93.75% of the data.
b. Determine the actual percentages of the given data values that fall in each of the intervals that
you calculated in part a. Also calculate the percentage of the data values that fall within one
standard deviation of the mean.
c. Do you think the lower endpoints provided by Chebyshev's theorem in part a are useful for this
problem? Explain your answer.
d. Suppose that the individual with the first number (54.4) in the fifth row of the data is a worka-
holic who actually worked 84.4 hours last week and not 54.4 hours. With this change now
x = 49.61 and s = 7.10. Recalculate the intervals for part a and the actual percentages for part b.
Did your percentages change a lot or a little?
e. How many standard deviations above the mean would you have to go to capture all 50 data val-
ues? What is the lower bound for the percentage of the data that should fall in the interval, ac-
cording to the Chebyshev theorem.
3.145 Refer to the women's golf scores in Exercise 3.137. It turns out that 117 was mistakenly entered.
Although this person still had the highest score among the 15 women, her score was not a mild or ex-
treme outlier according to the box-and-whisker plot, nor was she tied for the highest score. What are the
possible scores that she could have shot?
3.146 Refer to Problem 5 of the Self-Review Test in Chapter 1, which featured the 10 biggest Nasdaq
losers of October 2008. Let x represent the October 2008 return. Calculate Xx/10. What, if anything, does
this number represent? Explain your answer.
Appendix 127
APPENDIX 3.1
A3.1.1 BASIC FORMULAS FOR THE VARIANCE AND STANDARD
DEVIATION FOR UNGROUPED DATA
Example 3-25 below illustrates how to use the basic formulas to calculate the variance and standard
deviation for ungrouped data. From Section 3.2.2, the basic formulas for variance for ungrouped data are
%{x - /j,) 2 X(x - xf
a~ = and s =
N n-\
where a 2 is the population variance and s 2 is the sample variance.
In either case, the standard deviation is obtained by taking the square root of the variance.
EXAMPLE 3-25 Refer to Example 3-12, where we used the short-cut formulas to compute the vari-
ance and standard deviation for the data on the 2008 market values of five international companies. Cal-
culate the variance and standard deviation for those data using the basic formula.
Solution Let x denote the 2008 market values (in billions of dollars) of these companies. Table 3.14
shows all the required calculations to find the variance and standard deviation.
Calculating the variance and
standard deviation for ungrouped
data using basic formulas.
Table 3.14
X
(x-x)
(x - xf
75
75 -
132.4 =
-57.4
3294.76
107
107 -
132.4 =
-25.4
645.16
271
271 -
132.4 =
138.6
19,209.96
138
138 -
132.4 =
5.6
31.36
71
71 -
132.4 =
-61.4
3769.96
tx = 662
S(jc - xf = 26,951.20
The following steps are performed to compute the variance and standard deviation.
Step 1. Find the mean as follows:
2* 662
x = — = = 132.4
n 5
Step 2. Calculate x — x, the deviation of each value of x from the mean. The results are shown in the
second column of Table 3.14.
Step 3. Square each of the deviations of x from x; that is, calculate each of the (x — xf values. These val-
ues are called the squared deviations, and they are recorded in the third column.
Step 4. Add all the squared deviations to obtain 2(x — xf; that is, sum all the values given in the third
column of Table 3.14. This gives
X(x - xf = 26,951.20
Step 5. Obtain the sample variance by dividing the sum of the squared deviations by n — 1. Thus
„ %(x - xf 26,951.20
s 2 = = = 6737.80
n — 1 5 — 1
Step 6. Obtain the sample standard deviation by taking the positive square root of the variance. Hence,
5 = V6737.80 = 82.0841 = $82.08 billion
128 Chapter 3 Numerical Descriptive Measures
A3.1.2 BASIC FORMULAS FOR THE VARIANCE AND
STANDARD DEVIATION FOR GROUPED DATA
Example 3-26 demonstrates how to use the basic formulas to calculate the variance and standard devia-
tion for grouped data. The basic formulas for these calculations are
X/(m- M ) 2 M™-x) 2
a = and s~ =
N n - I
where cr is the population variance, s 2 is the sample variance, m is the midpoint of a class, and / is the
frequency of a class.
In either case, the standard deviation is obtained by taking the square root of the variance.
EXAMPLE 3-26 In Example 3-17, we used the short-cut formula to compute the variance and standard
deviation for the data on the numbers of orders received each day during the past 50 days at the office of
a mail-order company. Calculate the variance and standard deviation for those data using the basic formula.
Solution All the required calculations to find the variance and standard deviation appear in Table 3.15.
Table 3.15
Number of
Orders
/
m
mf
m — X
(m - x) 2
f(m - x) 2
10-12
4
11
44
-5.64
31.8096
127.2384
13-15
12
14
168
-2.64
6.9696
83.6352
16-18
20
17
340
.36
.1296
2.5920
19-21
14
20
280
3.36
11.2896
158.0544
n = 50
Xmf = 832
£/(« " x) 2
= 371.5200
Calculating the variance and
standard deviation for grouped
data using basic formulas.
The following steps are performed to compute the variance and standard deviation using the basic formula.
Step 1. Find the midpoint of each class. Multiply the corresponding values of m and/ Find %mf. From
Table 3.15, tmf = 832.
Step 2. Find the mean as follows:
x = Xmf/n = 832/50 = 16.64
Step 3. Calculate m — x, the deviation of each value of m from the mean. These calculations are done
in the fifth column of Table 3.15.
Step 4. Square each of the deviations m — x; that is, calculate each of the (m — x) values. These are
called squared deviations, and they are recorded in the sixth column.
Step 5. Multiply the squared deviations by the corresponding frequencies (see the seventh column of
Table 3.15). Adding the values of the seventh column, we obtain
%f(m - xf = 371.5200
Step 6. Obtain the sample variance by dividing Xf(m — x) 2 by n — 1. Thus,
S/(m - x) 2 371.5200
.v 2 = = = 7.5820
n - 1 50-1
Step 7. Obtain the standard deviation by taking the positive square root of the variance.
s = Vs 2 = V7.5820 = 2.75 orders ■
Self-Review Test
1. The value of the middle term in a ranked data set is called the
a. mean b. median c. mode
2. Which of the following summary measures is/are influenced by extreme values?
a. mean b. median c. mode d. range
Self-Review Test 129
3. Which of the following summary measures can be calculated for qualitative data?
a. mean b. median c. mode
4. Which of the following can have more than one value?
a. mean b. median c. mode
5. Which of the following is obtained by taking the difference between the largest and the smallest val-
ues of a data set?
a. variance b. range c. mean
6. Which of the following is the mean of the squared deviations of x values from the mean?
a. standard deviation b. population variance c. sample variance
7. The values of the variance and standard deviation are
a. never negative b. always positive c. never zero
8. A summary measure calculated for the population data is called
a. a population parameter b. a sample statistic c. an outlier
9. A summary measure calculated for the sample data is called a
a. population parameter b. sample statistic c. box-and-whisker plot
10. Chebyshev's theorem can be applied to
a. any distribution b. bell-shaped distributions only c. skewed distributions only
11. The empirical rule can be applied to
a. any distribution b. bell-shaped distributions only c. skewed distributions only
12. The first quartile is a value in a ranked data set such that about
a. 75% of the values are smaller and about 25% are larger than this value
b. 50% of the values are smaller and about 50% are larger than this value
c. 25% of the values are smaller and about 75% are larger than this value
13. The third quartile is a value in a ranked data set such that about
a. 75% of the values are smaller and about 25% are larger than this value
b. 50% of the values are smaller and about 50% are larger than this value
c. 25% of the values are smaller and about 75% are larger than this value
14. The 75th percentile is a value in a ranked data set such that about
a. 75% of the values are smaller and about 25% are larger than this value
b. 25% of the values are smaller and about 75% are larger than this value
15. The following data give the numbers of times 10 persons used their credit cards during the past 3
months.
9 6 28 14 2 18 7 3 16 6
Calculate the mean, median, mode, range, variance, and standard deviation.
16. The mean, as a measure of central tendency, has the disadvantage of being influenced by extreme
values. Illustrate this point with an example.
17. The range, as a measure of spread, has the disadvantage of being influenced by extreme values.
Illustrate this point with an example.
18. When is the value of the standard deviation for a data set zero? Give one example of such a data set.
Calculate the standard deviation for that data set to show that it is zero.
19. The following table gives the frequency distribution of the numbers of computers sold during the past
25 weeks at a computer store.
Computers Sold
Frequency
4 to 9
2
10 to 15
4
16 to 21
10
22 to 27
6
28 to 33
3
a. What does the frequency column in the table represent?
b. Calculate the mean, variance, and standard deviation.
130
Chapter 3 Numerical Descriptive Measures
20. The cars owned by all people living in a city are, on average, 7.3 years old with a standard deviation
of 2.2 years.
a. Using Chebyshev's theorem, find at least what percentage of the cars in this city are
i. 1.8 to 12.8 years old ii. .7 to 13.9 years old
b. Using Chebyshev's theorem, find the interval that contains the ages of at least 75% of the cars
owned by all people in this city.
21. The ages of cars owned by all people living in a city have a bell-shaped distribution with a mean of
7.3 years and a standard deviation of 2.2 years.
a. Using the empirical rule, find the percentage of cars in this city that are
i. 5.1 to 9.5 years old ii. .7 to 13.9 years old
b. Using the empirical rule, find the interval that contains the ages of 95% of the cars owned by
all people in this city.
22. The following data give the number of times the metal detector was set off by passengers at a small
airport during 15 consecutive half-hour periods on February 1, 2009.
7 2 1 21 3 8 1
1 53 5 1 42 01 1 14
a. Calculate the three quartiles and the interquartile range. Where does the value of 4 lie in rela-
tion to these quartiles?
b. Find the (approximate) value of the 60th percentile. Give a brief interpretation of this value.
c. Calculate the percentile rank of 12. Give a brief interpretation of this value.
23. Make a box-and-whisker plot for the data on the number of times passengers set off the airport metal
detector given in Problem 22. Comment on the skewness of this data set.
*24. The mean weekly wages of a sample of 15 employees of a company are $1035. The mean weekly
wages of a sample of 20 employees of another company are $1090. Find the combined mean for these
35 employees.
*25. The mean GPA of five students is 3.21. The GPAs of four of these five students are, respectively,
3.85, 2.67, 3.45, and 2.91. Find the GPA of the fifth student.
*26. The following are the prices (in thousands of dollars) of 10 houses sold recently in a city:
479 366 238 207 287 349 293 2534 463 538
Calculate the 10% trimmed mean for this data set. Do you think the 10% trimmed mean is a better sum-
mary measure than the (simple) mean (i.e., the mean of all 10 values) for these data? Briefly explain why
or why not.
*27. Consider the following two data sets.
Data Set I: 8 16 20 35
Data Set II: 5 13 17 32
Note that each value of the second data set is obtained by subtracting 3 from the corresponding value of
the first data set.
a. Calculate the mean for each of these two data sets. Comment on the relationship between the
two means.
b. Calculate the standard deviation for each of these two data sets. Comment on the relationship
between the two standard deviations.
Mini-Projects
■ MINI-PROJECT 3-1
Refer to the data you collected for Mini-Project 1-1 of Chapter 1 and analyzed graphically in Mini-Project 2-1
of Chapter 2. Write a report summarizing those data. This report should include answers to at least the
following questions.
a. Calculate the summary measures (mean, standard deviation, five-number summary, interquartile
range) for the variable you graphed in Mini-Project 2-1. Do this for the entire data set, as well as
for the different groups formed by the categorical variable that you used to divide the data set in
Mini-Project 2-1.
Decide for Yourself
b. Are the summary measures for the various groups similar to those for the entire data set? If not,
which ones differ and how do they differ? Make the same comparisons among the summary
measures for various groups. Do the groups have similar levels of variability? Explain how you
can determine this from the graphs that you created in Mini-Project 2—1.
C. Draw a box-and-whisker plot for the entire data set. Also draw side-by-side box-and-whisker plots
for the various groups. Are there any outliers? If so, are there any values that are outliers in any
of the groups but not in the entire data set? Does the plot show any skewness?
d. Discuss which measures for the center and spread would be more appropriate to use to describe
your data set. Also, discuss your reasons for using those measures.
■ MINI-PROJECT 3-2
You are employed as a statistician for a company that makes household products, which are sold by part-
time salespersons who work during their spare time. The company has four salespersons employed in a
small town. Let us denote these salespersons by A, B, C, and D. The sales records (in dollars) for the past
6 weeks for these four salespersons are shown in the following table.
Week
A
B
C
D
1
1774
2205
1330
1402
2
1808
1507
1295
1665
3
1890
2352
1502
1530
4
1932
1939
1104
1826
5
1855
2052
1189
1703
6
1726
1630
1441
1498
Your supervisor has asked you to prepare a brief report comparing the sales volumes and the consistency
of sales of these four salespersons. Use the mean sales for each salesperson to compare the sales volumes,
and then choose an appropriate statistical measure to compare the consistency of sales. Make the calcu-
lations and write a report.
DECIDE FOR YOURSELF
Deciding Where to Live
By the time you get to college, you must have heard it over and over
again: "A picture is worth a thousand words." Now we have pictures
and numbers discussed in Chapters 2 and 3, respectively. Why both?
Well, although each one of them acts as a summary of a data set, it
is a combination of the pictures and numbers that tells a big part of
the story without having to look at the entire data set. Suppose that
you ask a realtor for information on the prices of homes in two dif-
ferent but comparable midwestern suburbs. Let us call these Suburbs
A and B. The realtor provides you with the following information
that is obtained from a random sample of 40 houses in each suburb:
a. The average price of homes in each of the two suburbs
b. The five-number summary of prices of homes in each
neighborhood
c. The histogram of the distribution of home prices for each suburb
All the information provided by the realtor is given in the following
two tables and two histograms shown in Figures 3.16 and 3.17. Note
that the second table gives the minimum and maximum prices of
homes (in thousands of dollars) for each suburb along with the val-
ues of Qj, median, and Q 3 (in thousands of dollars).
Suburb
A
B
Average Price (in thousands of dollars)
221.9
220.03
Minimum Q l Median
Q 3
Maximum
Suburb A 151.0 175.5 188.0
Suburb B 187.0 210.0 222.5
199.5
228.0
587.0
250.0
Before you decide which suburb you should buy the house in,
answer the following questions:
1. Examine the summary statistics and graphs given here.
2. Explain how the information given here can help you to make a
decision about the suburb where you should look for a house to buy.
3. Explain how and why you might be misled by simply looking at the
average prices if you are looking to spend less money to buy a house.
4. Is there any information about the suburbs not given here that you
will like to obtain before making a decision about the suburb where
you should buy a house?
132 Chapter 3 Numerical Descriptive Measures
300 400
Prices of Homes
18 .~
16-
14-
12-
i io-
3
Q" „
a 8 -
£
6-
4-
2-
0--
2D0 220
Prices of Homes
Figure 3.16 Histogram of prices of homes in Suburb A.
Figure 3.17 Histogram of prices of homes in Suburb B.
ECHNOLOGY
INSTRUCTION
1-Uar Stats
x=6. 333333333
Ix=41
2x2=377
Sx=4. 400757511
(tx=4. 017323598
in=6
Screen 3.1
1-Uar Stats
tn=6
ninX=2
Qi=3
Med=6
Gh = ll
naxX=13
Screen 3.2
Numerical Descriptive Measures
1. To calculate the sample statistics (e.g., mean, standard deviation, and five -number sum-
mary), first enter your data into a list such as LI, then select STAT >CALC >1-Var
Stats, and press Enter. Access the name of your list by pressing 2nd >STAT and scroll-
ing through the list of names until you get to your list name. Press ENTER. You will
obtain the output shown in Screens 3.1 and 3.2.
Screen 3.1 shows, in this order, the sample mean, the sum of the data values, the
sum of the squared data values, the sample standard deviation, the value of the popula-
tion standard deviation (you will use this only when your data constitute a census
instead of a sample), and the number of data values (e.g., the sample or population
size). Pressing the downward arrow key will show the five-number summary, which is
shown in Screen 3.2.
2. Constructing a box-and-whisker plot is similar to constructing
a histogram. First enter your data into a list such as LI, then
select STAT PLOT and go into one of the three plots. Make
sure the plot is turned on. For the type, select the second row,
first column (this boxplot will display outliers, if there are
any). Enter the name of your list for XList. Select ZOOM>9
to display the plot as shown in Screen 3.3.
Screen 3.3
To find the sample statistics (e.g., the mean, standard deviation, and five-number summary),
first enter the given data in a column such as CI, and then select Stat >Basic Statistics >
Display Descriptive Statistics. In the dialog box you obtain, enter the name of the column
where your data are stored in the Variables box as shown in Screen 3.4. Click the
Statistics button in this dialog box and choose the summary measures you want to
Technology Instruction 133
Screen 3.4
Display Descriptive Statistics
CI
data
Variables:
data
By variables (optional):
Select
Statisti
Help
OK
cs... | Graphs...
Cancel
calculate in the new dialog box as shown in Screen 3.5. Click OK in both dialog boxes.
The output will appear in the Session window, which is shown in Screen 3.6.
Screen 3.5
Descriptive Statistics - Statistics
m
F Mean
V Trimmed mean
F N nonmissing
I - SE of mean
r Sum
r N
missing
W Standard deviation
F Minimum
F N total
V Variance
F Maximum
V Cumulative N
I - Coefficient of variation
I - Range
f - Percent
V Cumulative percent
F First quartile
V Sum of squares
F Median
I - Skewness
F Third quartile
I - Kurtosis
F Interquartile range
r MSSD
Help
OK
Cancel
IS Session
Descriptive Statistics:
data
Total
Variable Count Mean
StDev Minimum
Ql Hedian
Q3
Maximum IQR
data 5 6.33
4.40 2.00
2.75 6.00
11.50
13.00 8.75
V
<
Screen 3.6
1 34 Chapter 3 N umerical Descriptive Measures
2. To create a box-and-whisker plot, enter the given data in a column such as CI, select
Graph >Boxplot >Simple, and click OK. In the dialog box you obtain, enter the name
of the column with data in the Graph Variables box (see Screen 3.7) and click OK. The
boxplot shown in Screen 3.8 will appear.
Screen 3.7
Boxplot - One Y, Simple
Graph variables:
Scale..
Labels..
Data View..
Multiple Graphs..
Data Options..
Help
OK
Cancel
Screen 3.8
Boxplot of data
Calculating Summary Statistics Using the Excel Analysis ToolPak Add-in
1. Click the Data tab. Click Data Analysis in the Analysis group. The Data Analysis menu
will open (see Screen 3.9).
2. Select Descriptive Statistics. Click OK. The Descriptive Statistics window will open (see
Screen 3.10). Click in the Input Range box. Select the range where your data are located.
(Note: the easiest way to do this is to highlight the data with your mouse.) Select Rows or
Columns to identify whether the data are grouped in rows or columns.
Technology Instruction 135
Data Analysis
Analysis Tools
Anova: Single Factor
Anova: Two-Factor With Replication
Anova: Two-Factor Without Replication
Correlation
Covariance
Descriptive Statistics
Exponential Smoothing
F-Test Two-Sample for Variances
Fourier Analysis
j-fistiQgram
OK
Cancel
Help
Yi
Screen 3.10
Descriptive Statistics
Input
Input Range:
Grouped By:
Labels in first row
$A$3:$A$8
Columns
O Rows
Output options
O Output Range:
© New Worksheet Ply:
New Workbook
[^1 Summary statistics
1 I Confidence Level for Mean:
□ Kth Largest:
□ Kth Smallest:
my output
%
?j[x
OK
Cancel
Help
3. Select where you want Excel to place the output. You can select a specific range in the
current spreadsheet, a new spreadsheet within the current Excel workbook, or a new Excel
workbook.
4. Click Summary Statistics. Click OK (see Screen 3.11 for an example of the output).
5. The Analysis ToolPak does not calculate the first and third quartiles. To do this, go to an
empty cell in the spreadsheet. Then
a. Type =quartile(
b. Select the range of data and then type a comma
C. Type 1 for the first quartile or 3 for the third quartile
d. Type a right parenthesis, and then press Enter.
6. To find the Mi percentile:
a. Type =percentile(
b. Select the range of data and then type a comma
C. Type the value of k
d. Type a right parenthesis, and then press Enter.
1 36 Chapter 3 Numerical Descriptive Measures
1
Column 1
2
1
Mean
6.S33333
4
Standard Error
1.796602
5
Median
6
6
Mode
#N/A
7
Standard Deviation
4.400753
S
Sample Variance
19.36667
9
Kurtosis
-1.54594
10
Skewness
0.4626S
n
Range
11
12
Minimum
2
13
Maximum
13
14
Sum
41
15
Count
6
V J
TECHNOLOGY ASSIGNMENTS
TA3.1 Refer to the subsample taken in the Technology Assignment TA2.3 of Chapter 2 from the sam-
ple data on the time taken to run the Manchester Road Race. Find the mean, median, range, and standard
deviation for those data.
TA3.2 Refer to the data on phone charges given in Data Set I. From that data set select the 4th value
and then select every 10th value after that (i.e., select the 4th, 14th, 24th, 34th, ... values). Such a sample
taken from a population is called a systematic random sample. Find the mean, median, standard deviation,
first quartile, and third quartile for the phone charges for this subsample.
TA3.3 Refer to Data Set I on the prices of various products in different cities across the country. Select
a subsample of the prices of regular unleaded gas for 40 cities. Find the mean, median, and standard de-
viation for the data of this subsample.
TA3.4 Refer to Data of TA3.3. Make a box-and-whisker plot for those data.
TA3.5 Refer to Data Set I on the prices of various products in different cities across the country. Make
a box-and-whisker plot for the data on the monthly telephone charges.
TA3.6 Refer to the data on the numbers of computer keyboards assembled at the Twentieth Century Elec-
tronics Company for a sample of 25 days given in Exercise 3.104. Prepare a box-and-whisker plot for
those data.
TA3.7 Refer to Data Set VII on the stocks included in the Standard & Poor's 100 Index. Calculate the
mean, median, standard deviation, and interquartile range for the data on the highest prices for the stocks
in each of the market sectors. Compare the values of the various statistics for different sectors. Create a
stacked dotplot of the highest prices for various sectors with each sector's data as one set of data. Explain
how the results of your comparisons can be seen in the dotplot.
TA3.8 Refer to Data Set III on the National Basketball Association. Calculate the mean, median, stan-
dard deviation, and interquartile range for the players' ages separately for each of the three primary posi-
tions (center, forward, and guard). Is there a position that tends to have younger players, on average, than
the other positions? Is there a position that tends to have less variability in the ages of the players?
TA3.9 Calculate the five-number summaries, the values of the upper and lower inner fences, and the val-
ues of the upper and lower outer fences for the data referred to in TA3.8. Create side-by-side boxplots for
the data on three primary positions. Using these boxplots, compare the shapes of the age distributions for
the three positions. Are there any outliers? If so, classify the outliers as being mild or extreme.
Probability
Do you always come to a complete stop when approaching a stop sign? Although we all know
that we should, many of us rarely make a complete stop at stop signs. In a survey conducted by
Consumer Reports, 17% of men and 13% of women indicated that they often just slow down before
proceeding through a stop sign. On the other hand, 42% of men and 56% of women surveyed said
that they always stop at a stop sign. (See Case Study 4-1.)
We often make statements about probability. For example, a weather forecaster may predict that there
is an 80% chance of rain tomorrow. A health news reporter may state that a smoker has a much greater
chance of getting cancer than does a nonsmoker. A college student may ask an instructor about the
chances of passing a course or getting an A if he or she did not do well on the midterm examination.
Probability, which measures the likelihood that an event will occur, is an important part of statistics.
It is the basis of inferential statistics, which will be introduced in later chapters. In inferential statistics,
we make decisions under conditions of uncertainty. Probability theory is used to evaluate the uncer-
tainty involved in those decisions. For example, estimating next year's sales for a company is based
on many assumptions, some of which may happen to be true and others may not. Probability theory
will help us make decisions under such conditions of imperfect information and uncertainty.
Combining probability and probability distributions (which are discussed in Chapters 5 through 7) with
descriptive statistics will help us make decisions about populations based on information obtained
from samples. This chapter presents the basic concepts of probability and the rules for computing
probability.
4.1 Experiment, Outcomes,
and Sample Space
4.2 Calculating Probability
4.3 Counting Rule
4.4 Marginal and
Conditional Probabilities
Case Study 4-1 Rolling Stops
4.5 Mutually Exclusive
Events
4.6 Independent versus
Dependent Events
4.7 Complementary Events
4.8 Intersection of Events
and the Multiplication
Rule
Case Study 4-2 Baseball
Players have "Slumps"
and "Streaks"
4.9 Union of Events and the
Addition Rule
137
138 Chapter 4 Probability
4.1 Experiment, Outcomes, and Sample Space
Quality control inspector Jack Cook of Tennis Products Company picks up a tennis ball from
the production line to check whether it is good or defective. Cook's act of inspecting a tennis
ball is an example of a statistical experiment. The result of his inspection will be that the ball
is either "good" or "defective." Each of these two observations is called an outcome (also called
a basic or final outcome) of the experiment, and these outcomes taken together constitute the
sample space for this experiment.
Definition
Experiment, Outcomes, and Sample Space An experiment is a process that, when performed,
results in one and only one of many observations. These observations are called the outcomes of
the experiment. The collection of all outcomes for an experiment is called a sample space.
A sample space is denoted by S. The sample space for the example of inspecting a tennis
ball is written as
S = {good, defective}
The elements of a sample space are called sample points.
Table 4.1 lists some examples of experiments, their outcomes, and their sample spaces.
Table 4.1 Examples of Experiments, Outcomes, and Sample Spaces
Experiment
Outcomes
Sample Space
Toss a coin once
Head, Tail
S = {Head, Tail)
Roll a die once
1, 2, 3, 4, 5, 6
S = {1, 2, 3, 4, 5, 6)
Toss a coin twice
HH, HT, TH, TT
S = {HH, HT, TH, TT}
Play lottery
Win, Lose
S = {Win, Lose)
Take a test
Pass, Fail
5 = {Pass, Fail}
Select a worker
Male, Female
S = {Male, Female)
The sample space for an experiment can also be illustrated by drawing either a Venn diagram
or a tree diagram. A Venn diagram is a picture (a closed geometric shape such as a rectangle,
a square, or a circle) that depicts all the possible outcomes for an experiment. In a tree diagram,
each outcome is represented by a branch of the tree. Venn and tree diagrams help us understand
probability concepts by presenting them visually. Examples 4-1 through 4-3 describe how to
draw these diagrams for statistical experiments.
■ EXAMPLE 4-1
Draw the Venn and tree diagrams for the experiment of tossing a coin once.
Solution This experiment has two possible outcomes: head and tail. Consequently, the sam-
ple space is given by
S = {H, T}, where H = Head and T = Tail
To draw a Venn diagram for this example, we draw a rectangle and mark two points
inside this rectangle that represent the two outcomes, Head and Tail. The rectangle is
labeled S because it represents the sample space (see Figure 4.1a). To draw a tree diagram,
Drawing Venn and tree
diagrams: one toss of a coin.
4.1 Experiment, Outcomes, and Sample Space 139
we draw two branches starting at the same point, one representing the head and the sec-
ond representing the tail. The two final outcomes are listed at the ends of the branches (see
Figure 4.1b).
Outcomes
Head H
Figure 4.T (a) Venn diagram and (b) tree diagram
for one toss of a coin.
(a)
(b)
■ EXAMPLE 4-2
Draw the Venn and tree diagrams for the experiment of tossing a coin twice.
Solution This experiment can be split into two parts: the first toss and the second toss. Sup-
pose that the first time the coin is tossed, we obtain a head. Then, on the second toss, we can
still obtain a head or a tail. This gives us two outcomes: HH (head on both tosses) and HT
(head on the first toss and tail on the second toss). Now suppose that we observe a tail on the
first toss. Again, either a head or a tail can occur on the second toss, giving the remaining two
outcomes: TH (tail on the first toss and head on the second toss) and TT (tail on both tosses).
Thus, the sample space for two tosses of a coin is
S = {HH, HT, TH, TT}
The Venn and tree diagrams are given in Figure 4.2. Both of these diagrams show the sample
space for this experiment.
Drawing Venn and tree
diagrams: two tosses of a coin.
HH»
• HT
TH*
• TT
(a)
First toss
Second Final
toss outcomes
H HH
(b)
Figure 4.2 (a) Venn diagram and (b) tree diagram for two tosses
of a coin.
■ EXAMPLE 4-3
Suppose we randomly select two workers from a company and observe whether the worker
selected each time is a man or a woman. Write all the outcomes for this experiment. Draw the
Venn and tree diagrams for this experiment.
Solution Let us denote the selection of a man by M and that of a woman by W. We can
compare the selection of two workers to two tosses of a coin. Just as each toss of a coin can
result in one of two outcomes, head or tail, each selection from the workers of this company
can result in one of two outcomes, man or woman. As we can see from the Venn and tree di-
agrams of Figure 4.3, there are four final outcomes: MM, MW, WM, WW. Hence, the sample
space is written as
Drawing Venn and tree
diagrams: two selections.
S = {MM, MW, WM, WW}
140 Chapter 4 Probability
Figure 4.3 (a) Venn diagram and (b) tree dia-
gram for selecting two workers.
S
4.1.1 Simple and Compound Events
An event consists of one or more of the outcomes of an experiment.
Definition
Event An event is a collection of one or more of the outcomes of an experiment.
An event may be a simple event or a compound event. A simple event is also called an el-
ementary event, and a compound event is also called a composite event.
Simple Event
Each of the final outcomes for an experiment is called a simple event. In other words, a sim-
ple event includes one and only one outcome. Usually, simple events are denoted by E x , E 2 , E 3 ,
and so forth. However, we can denote them by any other letter — that is, by A, B, C, and so forth.
Definition
Simple Event An event that includes one and only one of the (final) outcomes for an experiment
is called a simple event and is usually denoted by E t ,
Example 4-4 describes simple events.
■ EXAMPLE 4-4
Reconsider Example 4-3 on selecting two workers from a company and observing whether
the worker selected each time is a man or a woman. Each of the final four outcomes (MM,
MW, WM, and WW) for this experiment is a simple event. These four events can be denoted
by E x , E 2 , E 3 , and E A , respectively. Thus,
Ei = (MM), E 2 = (MW), E 3 = (WM), and E 4 = (WW) ■
Compound Event
A compound event consists of more than one outcome.
Definition
Compound Event A compound event is a collection of more than one outcome for an experiment.
Compound events are denoted by A, B, C, D,... or by A x , A 2 , A 3 ,..., B lt B 2 , B 3 ,..., and so
forth. Examples 4—5 and 4-6 describe compound events.
MM •
• MW
WM •
• WW
(a)
First
selection
Second Final
selection outcomes
MM
WW
(b)
Illustrating simple events.
4.1 Experiment, Outcomes, and Sample Space 141
■ EXAMPLE 4-5
Reconsider Example 4-3 on selecting two workers from a company and observing whether the
worker selected each time is a man or a woman. Let A be the event that at most one man is se-
lected. Event A will occur if either no man or one man is selected. Hence, the event A is given by
A = {MW, WM, WW}
Because event A contains more than one outcome, it is a compound event. The Venn diagram
in Figure 4.4 gives a graphic presentation of compound event A. M
■ EXAMPLE 4-6
In a group of people, some are in favor of genetic engineering and others are against it. Two
persons are selected at random from this group and asked whether they are in favor of or
against genetic engineering. How many distinct outcomes are possible? Draw a Venn diagram
and a tree diagram for this experiment. List all the outcomes included in each of the follow-
ing events and state whether they are simple or compound events.
(a) Both persons are in favor of genetic engineering.
(b) At most one person is against genetic engineering.
(c) Exactly one person is in favor of genetic engineering.
Solution Let
F = a person is in favor of genetic engineering
A = a person is against genetic engineering
This experiment has the following four outcomes:
FF = both persons are in favor of genetic engineering
FA = the first person is in favor and the second is against
AF = the first person is against and the second is in favor
AA = both persons are against genetic engineering
The Venn and tree diagrams in Figure 4.5 show these four outcomes.
Illustrating a compound event-
two selections.
MM
/" \ A
•
/ MW \
I WM WW J
\ • • /
Figure 4.4 Venn diagram for
event A.
Illustrating simple and
compound events: two
selections.
Venn diagram
FF*
• FA
AF*
•AA
Tree diagram Figure 4.5 Venn and tree
First 1 Second Final diagrams,
person i person outcomes
F FF
(a)
(b)
(a) The event "both persons are in favor of genetic engineering" will occur if FF is ob-
tained. Thus,
Both persons are in favor of genetic engineering = {FF}
Because this event includes only one of the final four outcomes, it is a simple event.
(b) The event "at most one person is against genetic engineering" will occur if either none
or one of the persons selected is against genetic engineering. Consequently,
At most one person is against genetic engineering = {FF, FA, AF}
Because this event includes more than one outcome, it is a compound event.
1 42 Chapter 4 Probability
(c) The event "exactly one person is in favor of genetic engineering" will occur if one of
the two persons selected is in favor and the other is against genetic engineering. Hence,
it includes the following two outcomes:
Exactly one person is in favor of genetic engineering = {FA, AF}
Because this event includes more than one outcome, it is a compound event. H
4.1
4.2
EXERCISES
CONCEPTS AND PROCEDURES
Define the following terms: experiment, outcome, sample space, simple event, and compound event.
List the simple events for each of the following statistical experiments in a sample space S.
a. One roll of a die b. Three tosses of a coin c. One toss of a coin and one roll of a die
4.3 A box contains three items that are labeled A, B, and C. Two items are selected at random (without
replacement) from this box. List all the possible outcomes for this experiment. Write the sample space S.
■ APPLICATIONS
4.4 Two students are randomly selected from a statistics class, and it is observed whether or not they
suffer from math anxiety. How many total outcomes are possible? Draw a tree diagram for this experi-
ment. Draw a Venn diagram.
4.5 In a group of adults, some have iPods, and others do not. If two adults are randomly selected from
this group, how many total outcomes are possible? Draw a tree diagram for this experiment.
4.6 A test contains two multiple-choice questions. If a student makes a random guess to answer each
question, how many outcomes are possible? Depict all these outcomes in a Venn diagram. Also draw a
tree diagram for this experiment. (Hint: Consider two outcomes for each question — either the answer is
correct or it is wrong.)
4.7 A box contains a certain number of computer parts, a few of which are defective. Two parts are se-
lected at random from this box and inspected to determine if they are good or defective. How many total
outcomes are possible? Draw a tree diagram for this experiment.
4.8 In a group of people, some are in favor of a tax increase on rich people to reduce the federal deficit
and others are against it. (Assume that there is no other outcome such as "no opinion" and "do not know")
Three persons are selected at random from this group and their opinions in favor or against raising such
taxes are noted. How many total outcomes are possible? Write these outcomes in a sample space S. Draw
a tree diagram for this experiment.
4.9 Draw a tree diagram for three tosses of a coin. List all outcomes for this experiment in a sample
space S.
4.10 Refer to Exercise 4.4. List all the outcomes included in each of the following events. Indicate which
are simple and which are compound events.
a. Both students suffer from math anxiety.
b. Exactly one student suffers from math anxiety.
c. The first student does not suffer and the second suffers from math anxiety.
d. None of the students suffers from math anxiety.
4.11 Refer to Exercise 4.5. List all the outcomes included in each of the following events. Indicate which
are simple and which are compound events.
a. One person has an iPod and the other does not.
b. At least one person has an iPod.
c. Not more than one person has an iPod.
d. The first person has an iPod and the second does not.
4.12 Refer to Exercise 4.6. List all the outcomes included in each of the following events and mention
which are simple and which are compound events.
a. Both answers are correct.
b. At most one answer is wrong.
c. The first answer is correct and the second is wrong.
d. Exactly one answer is wrong.
4.2 Calculating Probability 143
4.13 Refer to Exercise 4.7. List all the outcomes included in each of the following events. Indicate which
are simple and which are compound events.
a. At least one part is good.
b. Exactly one part is defective.
c. The first part is good and the second is defective.
d. At most one part is good.
4.14 Refer to Exercise 4.8. List all the outcomes included in each of the following events and mention
which are simple and which are compound events.
a. At most one person is against a tax increase on rich people.
b. Exactly two persons are in favor of a tax increase on rich people.
c. At least one person is against a tax increase on rich people.
d. More than one person is against a tax increase on rich people.
4.2 Calculating Probability
Probability, which gives the likelihood of occurrence of an event, is denoted by P. The prob-
ability that a simple event E t will occur is denoted by P(Ej), and the probability that a com-
pound event A will occur is denoted by P(A).
Definition
Probability Probability is a numerical measure of the likelihood that a specific event will occur.
1. The probability of an event always lies in the range to 1. M Two Properties of Probability
Whether it is a simple or a compound event, the probability of an event is never less than or
greater than 1. Using mathematical notation, we can write this property as follows.
First Property of Probability
< P(E,) < 1
' P(A) ' 1
An event that cannot occur has zero probability; such an event is called an impossible event.
An event that is certain to occur has a probability equal to 1 and is called a sure event. That is,
For an impossible event M: P(M) =
For a sure event C: P(C) = 1
2. The sum of the probabilities of all simple events (or final outcomes) for an experiment,
denoted by tPiE,), is always 1.
Second Property of Probability For an experiment,
= P{E,) + P(E 2 ) + P(E 3 ) + ■■■=!
From this property, for the experiment of one toss of a coin,
P(H) + P{T) = 1
For the experiment of two tosses of a coin,
P(HH) + P(HT) + P(TH) + P(TT) = 1
For one game of football by a professional team,
P (win) + P(loss) + P (tie) = 1
1 44 Chapter 4 Probability
4.2.1 Three Conceptual Approaches to Probability
The three conceptual approaches to probability are (1) classical probability, (2) the relative fre-
quency concept of probability, and (3) the subjective probability concept. These three concepts
are explained next.
Classical Probability
Many times, various outcomes for an experiment may have the same probability of occurrence.
Such outcomes are called equally likely outcomes. The classical probability rule is applied to
compute the probabilities of events for an experiment for which all outcomes are equally likely.
Definition
Equally Likely Outcomes Two or more outcomes (or events) that have the same probability of
occurrence are said to be equally likely outcomes (or events).
According to the classical probability rule, the probability of a simple event is equal to 1
divided by the total number of outcomes for the experiment. This is obvious because the sum
of the probabilities of all final outcomes for an experiment is 1, and all the final outcomes are
equally likely. In contrast, the probability of a compound event A is equal to the number of out-
comes favorable to event A divided by the total number of outcomes for the experiment.
Find Probability
1
Total number of outcomes for the experiment
Number of outcomes favorable to A
Total number of outcomes for the experiment
Examples 4-7 through 4—9 illustrate how probabilities of events are calculated using the
classical probability rule.
Classical Probability Rule to
P(E l ) -
P(A)-
Calculating the probability
of a simple event.
I EXAMPLE 4-7
Find the probability of obtaining a head and the probability of obtaining a tail for one toss of
a coin.
Solution The two outcomes, head and tail, are equally likely outcomes. Therefore,
Similarly,
1 1
f(head) = = - = .50
Total number of outcomes 2
P(tail) = | = .50
Calculating the probability
of a compound event.
■ EXAMPLE 4-8
Find the probability of obtaining an even number in one roll of a die.
'If the final answer for the probability of an event does not terminate within three decimal places, usually it is rounded
to four decimal places.
4.2 Calculating Probability 145
Solution This experiment has a total of six outcomes: 1, 2, 3, 4, 5, and 6. All these out-
comes are equally likely. Let A be an event that an even number is observed on the die. Event
A includes three outcomes: 2, 4, and 6; that is,
A = {2, 4, 6}
If any one of these three numbers is obtained, event A is said to occur. Hence,
Number of outcomes included in A 3
P(A)
Total number of outcomes
.50
■ EXAMPLE 4-9
In a group of 500 women, 120 have played golf at least once. Suppose one of these 500 women
is randomly selected. What is the probability that she has played golf at least once?
Solution Because the selection is to be made randomly, each of the 500 women has the
same probability of being selected. Consequently this experiment has a total of 500 equally
likely outcomes. One hundred twenty of these 500 outcomes are included in the event that the
selected woman has played golf at least once. Hence,
120
500
.24
^(selected woman has played golf at least once)
Relative Frequency Concept of Probability
Suppose we want to calculate the following probabilities:
1. The probability that the next car that comes out of an auto factory is a "lemon"
2. The probability that a randomly selected family owns a home
3. The probability that a randomly selected woman has never smoked
4. The probability that an 80-year-old person will live for at least 1 more year
5. The probability that the tossing of an unbalanced coin will result in a head
6. The probability that a randomly selected person owns a sport-utility vehicle (SUV)
These probabilities cannot be computed using the classical probability rule because the various out-
comes for the corresponding experiments are not equally likely. For example, the next car manu-
factured at an auto factory may or may not be a lemon. The two outcomes, "it is a lemon" and "it
is not a lemon," are not equally likely. If they were, then (approximately) half the cars manufac-
tured by this company would be lemons, and this might prove disastrous to the survival of the firm.
Although the various outcomes for each of these experiments are not equally likely, each of
these experiments can be performed again and again to generate data. In such cases, to calculate
probabilities, we either use past data or generate new data by performing the experiment a large
number of times. The relative frequency of an event is used as an approximation for the probabil-
ity of that event. This method of assigning a probability to an event is called the relative frequency
concept of probability. Because relative frequencies are determined by performing an experiment,
the probabilities calculated using relative frequencies may change almost each time an experiment
is repeated. For example, every time a new sample of 500 cars is selected from the production line
of an auto factory, the number of lemons in those 500 cars is expected to be different. However,
the variation in the percentage of lemons will be small if the sample size is large. Note that if we
are considering the population, the relative frequency will give an exact probability.
Using Relative Frequency as an Approximation of Probability If an experiment is repeated n times
and an event A is observed /times, then, according to the relative frequency concept of probability,
Calculating the probability
of a compound event.
P(A)
/
Examples 4-10 and 4-11 illustrate how the probabilities of events are approximated using
the relative frequencies.
146 Chapter 4 Probability
■ EXAMPLE 4-10
Ten of the 500 randomly selected cars manufactured at a certain auto factory are found to be
lemons. Assuming that the lemons are manufactured randomly, what is the probability that the
next car manufactured at this auto factory is a lemon?
Solution Let n denote the total number of cars in the sample and / the number of lemons
in n. Then,
n = 500 and / = 10
Using the relative frequency concept of probability, we obtain
Pfnext car is a lemon) = — = = .02
This probability is actually the relative frequency of lemons in 500 cars. Table 4.2 lists the
frequency and relative frequency distributions for this example.
Table 4.2 Frequency and Relative Frequency
Distributions for the Sample of Cars
Car
/
Relative Frequency
Good
490
490/500 = .98
Lemon
10
10/500 = .02
n = 500
Sum = 1.00
The column of relative frequencies in Table 4.2 is used as the column of approximate prob-
abilities. Thus, from the relative frequency column,
^(next car is a lemon) = .02
^(next car is a good car) = .98 ^
Note that relative frequencies are not exact probabilities but are approximate proba-
bilities unless they are based on a census. However, if the experiment is repeated again and
again, this approximate probability of an outcome obtained from the relative frequency will
approach the actual probability of that outcome. This is called the Law of Large Numbers.
Definition
Law of Large Numbers If an experiment is repeated again and again, the probability of an event
obtained from the relative frequency approaches the actual or theoretical probability.
■ EXAMPLE 4-11
Allison wants to determine the probability that a randomly selected family from New York
State owns a home. How can she determine this probability?
Solution There are two outcomes for a randomly selected family from New York State: "This
family owns a home" and "this family does not own a home." These two events are not equally
likely. (Note that these two outcomes will be equally likely if exactly half of the families in
New York State own homes and exactly half do not own homes.) Hence, the classical proba-
bility rule cannot be applied. However, we can repeat this experiment again and again. In other
words, we can select a sample of families from New York State and observe whether or not
each of them owns a home. Hence, we will use the relative frequency approach to probability.
Approximating probability by
relative frequency: sample data.
Approximating probability by
relative frequency.
4.2 Calculating Probability 147
Suppose Allison selects a random sample of 1000 families from New York State and ob-
serves that 730 of them own homes and 270 do not own homes. Then,
n = sample size = 1000
/ = number of families who own homes = 730
Consequently,
/ 730
P(a randomly selected family owns a home) = — = ^qqq =
Again, note that .730 is just an approximation of the probability that a randomly selected
family from New York State owns a home. Every time Allison repeats this experiment she
may obtain a different probability for this event. However, because the sample size (n = 1000)
in this example is large, the variation is expected to be very small. I
Subjective Probability
Many times we face experiments that neither have equally likely outcomes nor can be repeated
to generate data. In such cases, we cannot compute the probabilities of events using the classi-
cal probability rule or the relative frequency concept. For example, consider the following prob-
abilities of events:
1. The probability that Carol, who is taking a statistics course, will earn an A in the course
2. The probability that the Dow Jones Industrial Average will be higher at the end of the next
trading day
3. The probability that the New York Giants will win the Super Bowl next season
4. The probability that Joe will lose the lawsuit he has filed against his landlord
Neither the classical probability rule nor the relative frequency concept of probability can
be applied to calculate probabilities for these examples. All these examples belong to experi-
ments that have neither equally likely outcomes nor the potential of being repeated. For exam-
ple, Carol, who is taking statistics, will take the test (or tests) only once, and based on that she
will either earn an A or not. The two events "she will earn an A" and "she will not earn an A"
are not equally likely. The probability assigned to an event in such cases is called subjective
probability. It is based on the individual's judgment, experience, information, and belief. Carol
may assign a high probability to the event that she will earn an A in statistics, whereas her in-
structor may assign a low probability to the same event.
Definition
Subjective Probability Subjective probability is the probability assigned to an event based on
subjective judgment, experience, information, and belief.
Subjective probability is assigned arbitrarily. It is usually influenced by the biases, prefer-
ences, and experience of the person assigning the probability.
EXERCISES
CONCEPTS AND PROCEDURES
4.15 Briefly explain the two properties of probability.
4.16 Briefly describe an impossible event and a sure event. What is the probability of the occurrence of
each of these two events?
4.17 Briefly explain the three approaches to probability. Give one example of each approach.
1 48 Chapter 4 Probability
4.18 Briefly explain for what kind of experiments we use the classical approach to calculate probabilities
of events and for what kind of experiments we use the relative frequency approach.
4.19 Which of the following values cannot be probabilities of events and why?
1/5 .97 -.55 1.56 5/3 0.0 -2/7 1.0
4.20 Which of the following values cannot be probabilities of events and why?
.46 2/3 -.09 1.42 .96 9/4 -1/4 .02
■ APPLICATIONS
4.21 Suppose a randomly selected passenger is about to go through the metal detector at JFK Airport in
New York City. Consider the following two outcomes: The passenger sets off the metal detector, and the
passenger does not set off the metal detector. Are these two outcomes equally likely? Explain why or why
not. If you are to find the probability of these two outcomes, would you use the classical approach or the
relative frequency approach? Explain why.
4.22 Thirty-two persons have applied for a security guard position with a company. Of them, 7 have
previous experience in this area and 25 do not. Suppose one applicant is selected at random. Consider
the following two events: This applicant has previous experience, and this applicant does not have pre-
vious experience. If you are to find the probabilities of these two events, would you use the classical
approach or the relative frequency approach? Explain why.
4.23 The president of a company has a hunch that there is a .80 probability that the company will be suc-
cessful in marketing a new brand of ice cream. Is this a case of classical, relative frequency, or subjective
probability? Explain why.
4.24 The coach of a college football team thinks that there is a .75 probability that the team will win the
national championship this year. Is this a case of classical, relative frequency, or subjective probability?
Explain why.
4.25 A hat contains 40 marbles. Of them, 18 are red and 22 are green. If one marble is randomly selected
out of this hat, what is the probability that this marble is
a. red? b. green?
4.26 A die is rolled once. What is the probability that
a. a number less than 5 is obtained?
b. a number 3 to 6 is obtained?
4.27 A random sample of 2000 adults showed that 1320 of them have shopped at least once on the Inter-
net. What is the (approximate) probability that a randomly selected adult has shopped on the Internet?
4.28 In a statistics class of 42 students, 28 have volunteered for community service in the past. Find the prob-
ability that a randomly selected student from this class has volunteered for community service in the past.
4.29 In a group of 50 car owners, 8 own hybrid cars. If one car owner is selected at random from this
group, what is the probability that this car owner owns a hybrid car?
4.30 Out of the 3000 families who live in a given apartment complex in New York City, 600 paid no in-
come tax last year. What is the probability that a randomly selected family from these 3000 families did
pay income tax last year?
4.31 A multiple-choice question on a test has five answers. If Dianne chooses one answer based on "pure
guess," what is the probability that her answer is
a. correct? b. wrong?
Do these two probabilities add up to 1.0? If yes, why?
4.32 There are 1265 eligible voters in a town, and 972 of them are registered to vote. If one eligible voter
is selected at random, what is the probability that this voter is
a. registered b. not registered?
Do these two probabilities add up to 1.0? If yes, why?
4.33 A company that plans to hire one new employee has prepared a final list of six candidates, all of whom
are equally qualified. Four of these six candidates are women. If the company decides to select at random
one person out of these six candidates, what is the probability that this person will be a woman? What is
the probability that this person will be a man? Do these two probabilities add up to 1.0? If yes, why?
4.34 A sample of 500 large companies showed that 120 of them offer free psychiatric help to their em-
ployees who suffer from psychological problems. If one company is selected at random from this sample,
4.3 Counting Rule
149
what is the probability that this company offers free psychiatric help to its employees who suffer from
psychological problems? What is the probability that this company does not offer free psychiatric help to its
employees who suffer from psychological problems? Do these two probabilities add up to 1.0? If yes, why?
4.35 A sample of 400 large companies showed that 130 of them offer free health fitness centers to their
employees on the company premises. If one company is selected at random from this sample, what is the
probability that this company offers a free health fitness center to its employees on the company prem-
ises? What is the probability that this company does not offer a free health fitness center to its employees
on the company premises? Do these two probabilities add up to 1.0? If yes, why?
4.36 In a large city, 15,000 workers lost their jobs last year. Of them, 7400 lost their jobs because their
companies closed down or moved, 4600 lost their jobs due to insufficient work, and the remainder lost
their jobs because their positions were abolished. If one of these 15,000 workers is selected at random,
find the probability that this worker lost his or her job
a. because the company closed down or moved
b. due to insufficient work
c. because the position was abolished
Do these probabilities add up to 1.0? If so, why?
4.37 A sample of 820 adults showed that 80 of them had no credit cards, 116 had one card each, 94 had
two cards each, 77 had three cards each, 43 had four cards each, and 410 had five or more cards each.
Write the frequency distribution table for the number of credit cards an adult possesses. Calculate the rel-
ative frequencies for all categories. Suppose one adult is randomly selected from these 820 adults. Find
the probability that this adult has
a. three credit cards b. five or more credit cards
4.38 In a sample of 500 families, 70 have a yearly income of less than $40,000, 220 have a yearly in-
come of $40,000 to $80,000, and the remaining families have a yearly income of more than $80,000. Write
the frequency distribution table for this problem. Calculate the relative frequencies for all classes. Sup-
pose one family is randomly selected from these 500 families. Find the probability that this family has a
yearly income of
a. less than $40,000 b. more than $80,000
4.39 Suppose you want to find the (approximate) probability that a randomly selected family from Los
Angeles earns more than $175,000 a year. How would you find this probability? What procedure would
you use? Explain briefly.
4.40 Suppose you have a loaded die and you want to find the (approximate) probabilities of different out-
comes for this die. How would you find these probabilities? What procedure would you use? Explain briefly.
4.3 Counting Rule
The experiments dealt with so far in this chapter have had only a few outcomes, which were easy
to list. However, for experiments with a large number of outcomes, it may not be easy to list all
outcomes. In such cases, we may use the counting rule to find the total number of outcomes.
Counting Rule to Find Total Outcomes If an experiment consists of three steps, and if the first step
can result in m outcomes, the second step in n outcomes, and the third step in k outcomes, then
Total outcomes for the experiment = m • n • k
The counting rule can easily be extended to apply to an experiment that has fewer or more
than three steps.
■ EXAMPLE 4-12
Suppose we toss a coin three times. This experiment has three steps: the first toss, the second
toss, and the third toss. Each step has two outcomes: a head and a tail. Thus,
rule: 3 steps.
Total outcomes for three tosses of a coin = 2X2X2 = 8
The eight outcomes for this experiment are HHH, HHT, HTH, HTT, THH, THT, TTH, and TTT. ■
1 50 Chapter 4 Probability
■ EXAMPLE 4-13
A prospective car buyer can choose between a fixed and a variable interest rate and can also
choose a payment period of 36 months, 48 months, or 60 months. How many total outcomes
are possible?
Solution This experiment is made up of two steps: choosing an interest rate and selecting
a loan payment period. There are two outcomes (a fixed or a variable interest rate) for the first
step and three outcomes (a payment period of 36 months, 48 months, or 60 months) for the
second step. Hence,
Total outcomes = 2X3 = 6 I
■ EXAMPLE 4-14
A National Football League team will play 16 games during a regular season. Each game can
result in one of three outcomes: a win, a loss, or a tie. The total possible outcomes for 16
games are calculated as follows:
Totalo utcomes = 3 , 3*3 , 3 , 3*3 , 3 , 3 , 3*3 , 3 , 3*3 , 3 , 3*3
= 3 16 = 43,046,721
One of the 43,046,721 possible outcomes is all 16 wins. I
4.4 Marginal and Conditional Probabilities
Suppose all 100 employees of a company were asked whether they are in favor of or against
paying high salaries to CEOs of U.S. companies. Table 4.3 gives a two-way classification of
the responses of these 100 employees.
Table 4.3 Two- Way Classification of Employee
Responses
In Favor
Against
Male
15
45
Female
4
36
Table 4.3 shows the distribution of 100 employees based on two variables or characteris-
tics: gender (male or female) and opinion (in favor or against). Such a table is called a contin-
gency table. In Table 4.3, each box that contains a number is called a cell. Notice that there are
four cells. Each cell gives the frequency for two characteristics. For example, 15 employees in
this group possess two characteristics: "male" and "in favor of paying high salaries to CEOs."
We can interpret the numbers in other cells the same way.
By adding the row totals and the column totals to Table 4.3, we write Table 4.4.
Table 4.4 Two-Way Classification of Employee
Responses with Totals
In Favor
Against
Total
Male
15
45
60
Female
4
36
40
Total
19
81
100
Suppose one employee is selected at random from these 100 employees. This employee
may be classified either on the basis of gender alone or on the basis of opinion. If only one
characteristic is considered at a time, the employee selected can be a male, a female, in favor,
or against. The probability of each of these four characteristics or events is called marginal
Applying the counting
rule: 2 steps.
Applying thecounting
rule: 16 steps.
4.4 Marginal and Conditional Probabilities 151
probability or simple probability. These probabilities are called marginal probabilities because
they are calculated by dividing the corresponding row margins (totals for the rows) or column
margins (totals for the columns) by the grand total.
Definition
Marginal Probability Marginal probability is the probability of a single event without consid-
eration of any other event. Marginal probability is also called simple probability.
For Table 4.4, the four marginal probabilities are calculated as follows:
Number of males 60
P(male) = = = .60
Total number of employees 100
As we can observe, the probability that a male will be selected is obtained by dividing the to-
tal of the row labeled "Male" (60) by the grand total (100). Similarly,
P(female) = 40/100 = .40
P(in favor) = 19/100 = .19
P(against) = 81/100 = .81
These four marginal probabilities are shown along the right side and along the bottom of
Table 4.5.
Table 4.5 Listing the Marginal Probabilities
In Favor Against
(A) (B) Total
Male(M) 15 45 60 p (M ) = 60/100 = .60
Female (F) 4 36 40_ P(F) = 40/100 = 40
Total 19 81 100
P(A) = 19/100 P(B) = 81/100
= .19 = .81
Now suppose that one employee is selected at random from these 100 employees. Further-
more, assume it is known that this (selected) employee is a male. In other words, the event that
the employee selected is a male has already occurred. What is the probability that the employee
selected is in favor of paying high salaries to CEOs? This probability is written as follows:
h Read as "given"
P(in favor I male)
The event whose probability * [_ This event has
is to be determined already occurred
This probability, P(in favor I male), is called the conditional probability of "in favor" given
that the event "male" has already happened. It is read as "the probability that the employee se-
lected is in favor given that this employee is a male."
Definition
Conditional Probability Conditional probability is the probability that an event will occur given
that another event has already occurred. If A and B are two events, then the conditional proba-
bility of A given B is written as
P(A | B)
and read as "the probability of A given that B has already occurred.'
152 Chapter 4 Probability
■ EXAMPLE 4-15
Compute the conditional probability P(in favor I male) for the data on 100 employees given
Calculating the conditional , ,
in Table 4.4.
probability: two-way table.
Solution The probability P(in favor I male) is the conditional probability that a randomly
selected employee is in favor given that this employee is a male. It is known that the event
"male" has already occurred. Based on the information that the employee selected is a male,
we can infer that the employee selected must be one of the 60 males and, hence, must belong
to the first row of Table 4.4. Therefore, we are concerned only with the first row of that table.
In Favor
Against
Total
Male
15
45
60
Males who Total number
are in favor of males
The required conditional probability is calculated as follows:
Number of males who are in favor 15
P(in favor male) = = — = .25
Total number of males 60
As we can observe from this computation of conditional probability, the total number of
males (the event that has already occurred) is written in the denominator and the number of
males who are in favor (the event whose probability we are to find) is written in the numera-
tor. Note that we are considering the row of the event that has already occurred. The tree di-
agram in Figure 4.6 illustrates this example.
Figure 4.6 Tree diagram.
Calculating the conditional
probability: two-way table.
■ EXAMPLE 4-16
For the data of Table 4.4, calculate the conditional probability that a randomly selected em-
ployee is a female given that this employee is in favor of paying high salaries to CEOs.
Solution We are to compute the probability P(female I in favor). Because it is known that
the employee selected is in favor of paying high salaries to CEOs, this employee must belong
to the first column (the column labeled "in favor") and must be one of the 19 employees who
are in favor.
In Favor
15
4 < Females who are in favor
19 < Total number of employees who are in favor
USA TODAY Snapshots®
ROLLING
STOPS
Rolling stops
Men are more likely than women to say they
don't come to a complete stop at stop signs.
How often they only slow down:
Men
Women
, Occasionally
°l f 3%A 30%
Occasionally
41%
_
Don't
know
1%
Source: Consumer Reports
National Research Center
survey n f i.nnn adnirs
Never
56%
1
By Anne R. Carey and Alejandro Gonzalez, USA TODAY
The accompanying chart shows the percentages of men and women who only slow down at stop signs
instead of coming to a complete stop. As the percentages in the chart show, when asked whether they only
slow down (instead of coming to a complete stop), 17% of men said often (that is, they often only slow
down and proceed instead of coming to a complete stop), 41% said occasionally (that is, they occasion-
ally only slow down but mostly stop at stop signs), and 42% said never (that is, they never slow down and
proceed but always make a complete stop at stop signs). Among women, the corresponding percentages
are 13%, 30%, and 56%, respectively. These percentages can be written as conditional probabilities as fol-
lows. Suppose one driver is selected at random. Then, given that this driver is a male, the probability is .42
that he makes a complete stop at stop signs. On the other hand, if this selected driver is a female, this
probability is .56. These probabilities can be written as:
^(selected driver makes a complete stop at stop signs | male) = .42
P(selected driver makes a complete stop at stop signs | female) = .56
... Source: USA TODAY, January 30, 2009.
Note that these are approximate probabilities because the percentages given in the chart are based on a copyright © 2009 USA TODAY. Chart
sample survey.
reproduced with permission.
.2105
Hence, the required probability is
Number of females who are in favor 4
P (female in favor) = = —
Total number of employees who are in favor 19
The tree diagram in Figure 4.7 illustrates this example.
Figure 4.7 Tree diagram.
We are to find the
probability of this event
Required probability
This event has
already occurred
153
Chapter 4 Probability
4.5 Mutually Exclusive Events
Events that cannot occur together are called mutually exclusive events. Such events do not
have any common outcomes. If two or more events are mutually exclusive, then at most one of
them will occur every time we repeat the experiment. Thus the occurrence of one event excludes
the occurrence of the other event or events.
Definition
Mutually Exclusive Events Events that cannot occur together are said to be mutually exclusive
events.
For any experiment, the final outcomes are always mutually exclusive because one and only
one of these outcomes is expected to occur in one repetition of the experiment. For example,
consider tossing a coin twice. This experiment has four outcomes: HH, HT, TH, and TT. These
outcomes are mutually exclusive because one and only one of them will occur when we toss
this coin twice.
■ EXAMPLE 4-17
Consider the following events for one roll of a die:
A = an even number is observed = {2, 4, 6}
B = an odd number is observed = {1, 3, 5}
C = a number less than 5 is observed = {1,2, 3, 4}
Are events A and B mutually exclusive? Are events A and C mutually exclusive?
Solution Figures 4.8 and 4.9 show the diagrams of events A and B and events A and C, re-
spectively.
Illustrating mutually
exclusive and mutually
nonexclusive events.
Figure 4.8 Mutually exclusive Figure 4.9 Mutually nonex-
events A and B. elusive events A and C.
As we can observe from the definitions of events A and B and from Figure 4.8, events A
and B have no common element. For one roll of a die, only one of the two events A and B
can happen. Hence, these are two mutually exclusive events. We can observe from the defini-
tions of events A and C and from Figure 4.9 that events A and C have two common outcomes:
2-spot and 4-spot. Thus, if we roll a die and obtain either a 2-spot or a 4-spot, then A and C
happen at the same time. Hence, events A and C are not mutually exclusive. H
■ EXAMPLE 4-18
Consider the following two events for a randomly selected adult:
Illustrating mutually
exclusive events. Y = this adult has shopped on the Internet at least once
N = this adult has never shopped on the Internet
Are events Y and N mutually exclusive?
4.6 Independent Versus Dependent Events 155
Solution Note that event Y consists of all adults who have shopped on the Internet at least
once, and event N includes all adults who have never shopped on the Internet. These two events
are illustrated in the Venn diagram in Figure 4.10.
As we can observe from the definitions of events Y and N and from Figure 4.10, events Y
and N have no common outcome. They represent two distinct sets of adults: the ones who have
shopped on the Internet at least once and the ones who have never shopped on the Internet.
Hence, these two events are mutually exclusive. M
4.6 Independent Versus Dependent Events
In the case of two independent events, the occurrence of one event does not change the prob-
ability of the occurrence of the other event.
Definition
Independent Events Two events are said to be independent if the occurrence of one does not
affect the probability of the occurrence of the other. In other words, A and B are independent
events if
either P(A | B) = P(A) or P(B \ A) = P(B)
It can be shown that if one of these two conditions is true, then the second will also be true,
and if one is not true, then the second will also not be true.
If the occurrence of one event affects the probability of the occurrence of the other event,
then the two events are said to be dependent events. In probability notation, the two events are
dependent if either P(A | B) + P(A) or P(B \ A) + P{B).
■ EXAMPLE 4-19
Refer to the information on 100 employees given in Table 4.4 in Section 4.4. Are events
"female (F)" and "in favor (A)" independent?
Solution Events F and A will be independent if
P(F) = P(F | A)
Otherwise they will be dependent.
Using the information given in Table 4.4, we compute the following two probabilities:
P(F) = 40/100 = .40 and P(F\A) = 4/19 = .2105
Because these two probabilities are not equal, the two events are dependent. Here, depend-
ence of events means that the respective percentages of males who are in favor of and against
paying high salaries to CEOs are different from the respective percentages of females who are
in favor and against.
In this example, the dependence of A and F can also be proved by showing that the prob-
abilities P(A) and P(A | F) are not equal. I
Illustrating two dependent
events: two-way table.
156 Chapter 4 Probability
■ EXAMPLE 4-20
A box contains a total of 100 CDs that were manufactured on two machines. Of them, 60 were
manufactured on Machine I. Of the total CDs, 15 are defective. Of the 60 CDs that were man-
ufactured on Machine I, 9 are defective. Let D be the event that a randomly selected CD is
defective, and let A be the event that a randomly selected CD was manufactured on Machine
I. Are events D and A independent?
independent events.
Solution From the given information,
P{D) = 15/100 = .15 and P{D \ A) = 9/60 = .15
Hence,
P(D) = P(D | A)
Consequently, the two events, D and A, are independent.
Independence, in this example, means that the probability of any CD being defective is the
same, .15, irrespective of the machine on which it is manufactured. In other words, the two
machines are producing the same percentage of defective CDs. For example, 9 of the 60 CDs
manufactured on Machine I are defective, and 6 of the 40 CDs manufactured on Machine II
are defective. Thus, for each of the two machines, 15% of the CDs produced are defective.
Using the given information, we can prepare Table 4.6. The numbers in the shaded cells are
given to us. The remaining numbers are calculated by doing some arithmetic manipulations.
Table 4.6 Two-Way Classification Table
Defective
Good
(D)
(G)
Total
Machine I (A)
51
m
Machine II (fl)
6
34
40
Total
m
85
[ioo1
Using this table, we can find the following probabilities:
P{D) = 15/100 = .15
P{D\A) = 9/60 = .15
Because these two probabilities are the same, the two events are independent. I
Two Important Observations ► We can make the following two important observations about mutually exclusive, independent,
and dependent events.
1. Two events are either mutually exclusive or independent. 2
a. Mutually exclusive events are always dependent.
b. Independent events are never mutually exclusive.
2. Dependent events may or may not be mutually exclusive.
4.7 Complementary Events
Two mutually exclusive events that taken together include all the outcomes for an experiment
are called complementary events. Note that two complementary events are always mutually
exclusive.
2 The exception to this rule occurs when at least one of the two events has a zero probability.
4.7 Complementary Events 157
Definition
Complementary Events The complement of event A, denoted by A and read as "A bar" or "A
complement," is the event that includes all the outcomes for an experiment that are not in A.
Events A and A are complements of each other. The Venn diagram in Figure 4.11 shows
the complementary events A and A.
Figure 4.T 1 Venn diagram of two complementary events.
Because two complementary events, taken together, include all the outcomes for an exper-
iment and because the sum of the probabilities of all outcomes is 1, it is obvious that
P(A) + P(A) = 1
From this equation, we can deduce that
P(A) = 1 - P(A) and P(A) = 1 - P(A)
Thus, if we know the probability of an event, we can find the probability of its comple-
mentary event by subtracting the given probability from 1 .
■ EXAMPLE 4-21
In a group of 2000 taxpayers, 400 have been audited by the IRS at least once. If one taxpayer
is randomly selected from this group, what are the two complementary events for this exper-
iment, and what are their probabilities?
Calculating probabilities of
complementary events.
Solution The two complementary events for this experiment are
A = the selected taxpayer has been audited by the IRS at least once
A = the selected taxpayer has never been audited by the IRS
Note that here event A includes the 400 taxpayers who have been audited by the IRS at least
once, and A includes the 1600 taxpayers who have never been audited by the IRS. Hence, the
probabilities of events A and A are
P(A) = 400/2000 = .20 and P(A) = 1600/2000 = .80
As we can observe, the sum of these two probabilities is one. Figure 4.12 shows a Venn diagram
for this example.
Figure 4.12 Venn diagram.
158 Chapter 4 Probability
Calculating probabilities of
complementary events.
■ EXAMPLE 4-22
In a group of 5000 adults, 3500 are in favor of stricter gun control laws, 1200 are against such
laws, and 300 have no opinion. One adult is randomly selected from this group. Let A be the
event that this adult is in favor of stricter gun control laws. What is the complementary event
of A? What are the probabilities of the two events?
Solution The two complementary events for this experiment are
A = the selected adult is in favor of stricter gun control laws
A = the selected adult is either against such laws or has no opinion
Note that here event A includes 1500 adults who are either against stricter gun control laws
or have no opinion. Also notice that events A and A are complements of each other. Because
3500 adults in the group favor stricter gun control laws and 1500 either are against stricter
gun control laws or have no opinion, the probabilities of events A and A are
P(A) = 3500/5000 = .70 and P(A) = 1500/5000 = .30
As we can observe, the sum of these two probabilities is 1 . Also, once we find P(A), we can
find the probability of P(A) as
P(A) = 1 - P(A) = 1 - .70 = .30
Figure 4.13 shows a Venn diagram for this example.
Figure 4.13 Venn diagram.
I
EXERCISES
CONCEPTS AND PROCEDURES
4.41 Briefly explain the difference between the marginal and conditional probabilities of events. Give one
example of each.
4.42 What is meant by two mutually exclusive events? Give one example of two mutually exclusive events
and another example of two mutually nonexclusive events.
4.43 Briefly explain the meaning of independent and dependent events. Suppose A and B are two events.
What formula can you use to prove whether A and B are independent or dependent?
4.44 What is the complement of an event? What is the sum of the probabilities of two complementary
events?
4.45 How many different outcomes are possible for four rolls of a die?
4.46 How many different outcomes are possible for 10 tosses of a coin?
4.47 A statistical experiment has eight equally likely outcomes that are denoted by 1,2, 3, 4, 5, 6, 7, and
8. Let event A = {2, 5, 7} and event B = {2, 4, 8}.
a. Are events A and B mutually exclusive events?
b. Are events A and B independent events?
c. What are the complements of events A and B, respectively, and their probabilities?
4.48 A statistical experiment has 10 equally likely outcomes that are denoted by 1, 2, 3, 4, 5, 6, 7, 8, 9,
and 10. Let event A = {3, 4, 6, 9} and event B = { 1, 2, 5).
a. Are events A and B mutually exclusive events?
b. Are events A and B independent events?
c. What are the complements of events A and S, respectively, and their probabilities?
4.7 Complementary Events 159
■ APPLICATIONS
4.49 A small ice cream shop has 10 flavors of ice cream and 5 kinds of toppings for its sundaes. How
many different selections of one flavor of ice cream and one kind of topping are possible?
4.50 A man just bought 4 suits, 8 shirts, and 12 ties. All of these suits, shirts, and ties coordinate with
each other. If he is to randomly select one suit, one shirt, and one tie to wear on a certain day, how many
different outcomes (selections) are possible?
4.51 A restaurant menu has four kinds of soups, eight kinds of main courses, five kinds of desserts, and
six kinds of drinks. If a customer randomly selects one item from each of these four categories, how many
different outcomes are possible?
4.52 A student is to select three classes for next semester. If this student decides to randomly select one
course from each of eight economics classes, six mathematics classes, and five computer classes, how
many different outcomes are possible?
4.53 Two thousand randomly selected adults were asked whether or not they have ever shopped on the
Internet. The following table gives a two-way classification of the responses.
Have Shopped
Have Never Shopped
Male
500
700
Female
300
500
a. If one adult is selected at random from these 2000 adults, find the probability that this adult
i. has never shopped on the Internet
ii. is a male
Hi. has shopped on the Internet given that this adult is a female
iv. is a male given that this adult has never shopped on the Internet
b. Are the events "male" and "female" mutually exclusive? What about the events "have shopped"
and "male?" Why or why not?
c. Are the events "female" and "have shopped" independent? Why or why not?
4.54 According to a March 2009 Gallup Poll (http://www.gallup.com/poll/117025/Support-Nuclear-
Energy-Inches-New-High.aspx), 71% of Republicans/Republican leaners and 52% of Democrats/Democrat
leaners favor the use of nuclear power. The survey consisted of 1012 American adults, approximately half
of whom were Republicans or Republican leaners. Suppose the following table gives the distribution of
responses of these 1012 adults.
Favor
Do not favor
Republicans/Republican leaners
381
128
Democrats/Democrat leaners
258
245
a. If one person is selected at random from this sample of 1012 U.S. adults, find the probability that
this person
i. does not favor the use of nuclear power
ii. is a Republican/Republican leaner
iii. favors the use of nuclear power given that the person is a Republican/Republican leaner
iv. is a Republican/Republican leaner given that the person does not favor the use of nuclear
power
b. Are the events favors and does not favor mutually exclusive? What about the events does not
favor and Republican/Republican leaner!
c. Are the events does not favor and Republican/Republican leaner independent? Why or why not?
4.55 Two thousand randomly selected adults were asked if they are in favor of or against cloning. The
following table gives the responses.
In Favor
Against
No Opinion
Male
395
405
100
Female
300
680
120
1 60 Chapter 4 Probability
a. If one person is selected at random from these 2000 adults, find the probability that this person is
i. in favor of cloning
ii. against cloning
iii. in favor of cloning given the person is a female
iv. a male given the person has no opinion
b. Are the events "male" and "in favor" mutually exclusive? What about the events "in favor" and
"against?" Why or why not?
c. Are the events "female" and "no opinion" independent? Why or why not?
4.56 Five hundred employees were selected from a city's large private companies, and they were asked
whether or not they have any retirement benefits provided by their companies. Based on this information,
the following two-way classification table was prepared.
Have Retirement Benefits
Yes No
Men
225 75
Women
150 50
a. If one employee is selected at random from these 500 employees, find the probability that this
employee
i. is a woman
ii. has retirement benefits
iii. has retirement benefits given the employee is a man
iv. is a woman given that she does not have retirement benefits
b. Are the events "man" and "yes" mutually exclusive? What about the events "yes" and "no?" Why
or why not?
c. Are the events "woman" and "yes" independent? Why or why not?
4.57 A consumer agency randomly selected 1700 flights for two major airlines, A and B. The following
table gives the two-way classification of these flights based on airline and arrival time. Note that "less than
30 minutes late" includes flights that arrived early or on time.
Less Than 30
30 Minutes to
More Than
Minutes Late
1 Hour Late
1 Hour Late
Airline A
429
390
92
Airline B
393
316
80
a. If one flight is selected at random from these 1700 flights, find the probability that this flight is
i. more than 1 hour late
ii. less than 30 minutes late
iii. a flight on airline A given that it is 30 minutes to 1 hour late
iv. more than 1 hour late given that it is a flight on airline B
b. Are the events "airline A" and "more than 1 hour late" mutually exclusive? What about the events
"less than 30 minutes late" and "more than 1 hour late?" Why or why not?
c. Are the events "airline B" and "30 minutes to 1 hour late" independent? Why or why not?
4.58 Two thousand randomly selected adults were asked if they think they are financially better off than
their parents. The following table gives the two-way classification of the responses based on the educa-
tion levels of the persons included in the survey and whether they are financially better off, the same as,
or worse off than their parents.
Less Than
High
More Than
High School
School
High School
Better off
140
450
420
Same as
60
250
110
Worse off
200
300
70
4.8 Intersection of Events and the Multiplication Rule
161
a. If one adult is selected at random from these 2000 adults, find the probability that this adult is
i. financially better off than his/her parents
ii. financially better off than his/her parents given he/she has less than high school education
iii. financially worse off than his/her parents given he/she has high school education
iv. financially the same as his/her parents given he/she has more than high school education
b. Are the events "better off" and "high school" mutually exclusive? What about the events "less
than high school" and "more than high school?" Why or why not?
c. Are the events "worse off" and "more than high school" independent? Why or why not?
4.59 There are a total of 160 practicing physicians in a city. Of them, 75 are female and 25 are pediatri-
cians. Of the 75 females, 20 are pediatricians. Are the events "female" and "pediatrician" independent?
Are they mutually exclusive? Explain why or why not.
4.60 Of a total of 100 CDs manufactured on two machines, 20 are defective. Sixty of the total CDs were
manufactured on Machine I, and 10 of these 60 are defective. Are the events "machine type" and "defec-
tive CDs" independent? (Note: Compare this exercise with Example 4-20.)
4.61 A company hired 30 new college graduates last week. Of these, 16 are female and 11 are business
majors. Of the 16 females, 9 are business majors. Are the events "female" and "business major" independ-
ent? Are they mutually exclusive? Explain why or why not.
4.62 Define the following two events for two tosses of a coin:
A = at least one head is obtained
B = both tails are obtained
a. Are A and B mutually exclusive events? Are they independent? Explain why or why not.
b. Are A and B complementary events? If yes, first calculate the probability of B and then calculate
the probability of A using the complementary event rule.
4.63 Let A be the event that a number less than 3 is obtained if we roll a die once. What is the probabil-
ity of A? What is the complementary event of A, and what is its probability?
4.64 According to a 2007 America's Families and Living Arrangements Census Bureau survey, 52.1 mil-
lion children lived with both of their parents in the same household, whereas 21.6 million lived with at
most one parent in the household. Assume that all U.S. children are included in this survey and that this
information is true for the current population. If one child is selected at random, what are the two com-
plementary events and their probabilities?
4.65 The probability that a randomly selected college student attended at least one major league base-
ball game last year is .12. What is the complementary event? What is the probability of this complementary
event?
4.8 Intersection of Events
and the Multiplication Rule
This section discusses the intersection of two events and the application of the multiplication
rule to compute the probability of the intersection of events.
4.8.1 Intersection of Events
The intersection of two events is given by the outcomes that are common to both events.
Definition
Intersection of Events Let A and B be two events defined in a sample space. The intersection
of A and B represents the collection of all outcomes that are common to both A and B and is
denoted by
A andB
162 Chapter 4 Probability
The intersection of events A and B is also denoted by either AflBor AB. Let
A = event that a family owns a DVD player
B = event that a family owns a digital camera
Figure 4.14 illustrates the intersection of events A and B. The shaded area in this figure gives
the intersection of events A and B, and it includes all the families who own both a DVD player
and a digital camera.
Figure 4.14 Intersection of events A and B.
Intersection of A and B
4.8.2 Multiplication Rule
Sometimes we may need to find the probability of two or more events happening together.
Definition
Joint Probability
The probability of the intersection of two events is called their joint probability.
It is written as
P(A and B)
The probability of the intersection of two events is obtained by multiplying the marginal
probability of one event by the conditional probability of the second event. This rule is called
the multiplication rule.
Multiplication Rule to Find Joint Probability The probability of the intersection of two events A
and B is
P(A and B) = P{A) P(B | A)
The joint probability of events A and B can also be denoted by P(A D B) or P(AB).
■ EXAMPLE 4-23
Table 4.7 gives the classification of all employees of a company by gender and college degree.
Calculating the joint probability ° r J r J J ° o o
of two events: two-way table. Table 4.7 Classification of Employees by Gender and Education
College
Not a College
Graduate
Graduate
(G)
(N)
Total
Male (M)
1
20
27
Female (F)
4
9
13
Total
11
29
40
If one of these employees is selected at random for membership on the employee-management
committee, what is the probability that this employee is a female and a college graduate?
Solution We are to calculate the probability of the intersection of the events "female" (denoted by
F) and "college graduate" (denoted by G). This probability may be computed using the formula
P(F and G) = P{F) P(G | F)
4.8 Intersection of Events and the Multiplication Rule
The shaded area in Figure 4.15 shows the intersection of the events "female" and "college
graduate."
College Figure 4.15 Intersection of events
Females graduates F and G.
Females and college
graduates
Notice that there are 13 females among 40 employees. Hence, the probability that a female
is selected is
P(F) = 13/40
To calculate the probability P(G I F), we know that F has already occurred. Consequently, the
employee selected is one of the 13 females. In the table, there are 4 college graduates among
13 female employees. Hence, the conditional probability of G given F is
P{G | F) = 4/13
The joint probability of F and G is
P(F and G) = P(F) P{G \ F) = (13/40)(4/13) = .100
Thus, the probability is .100 that a randomly selected employee is a female and a college graduate.
The probability in this example can also be calculated without using the multiplication rule.
As we can notice from Figure 4.15 and from the table, 4 employees out of a total of 40 are
female and college graduates. Hence, if any of these 4 employees is selected, the events "fe-
male" and "college graduate" both happen. Therefore, the required probability is
P(F and G) = 4/40 = .100 ■
We can compute three other joint probabilities for the table in Example 4-23 as follows:
P(M and G) = P(M) P(G \ M) = (27/40)(7/27) = .175
P(MandA0 = P{M) P{N \ M) = (27/40)(20/27) = .500
P{F and N) = P(F) P(N \ F) = (13/40)(9/13) = .225
The tree diagram in Figure 4.16 shows all four joint probabilities for this example. The joint
probability of F and G is highlighted.
1 64 Chapter 4 Probability
■ EXAMPLE 4-24
A box contains 20 DVDs, 4 of which are defective. If two DVDs are selected at random (with-
out replacement) from this box, what is the probability that both are defective?
Solution Let us define the following events for this experiment:
G x = event that the first DVD selected is good
D l = event that the first DVD selected is defective
G 2 = event that the second DVD selected is good
D 2 = event that the second DVD selected is defective
We are to calculate the joint probability of D x and D 2 , which is given by
P(D 1 and D 2 ) = P(D l )P(D 2 1 D\)
As we know, there are 4 defective DVDs in 20. Consequently, the probability of selecting
a defective DVD at the first selection is
P(D X ) = A/20
To calculate the probability P(D 2 \ D x ), we know that the first DVD selected is defective be-
cause D x has already occurred. Because the selections are made without replacement, there
are 19 total DVDs, and 3 of them are defective at the time of the second selection. Therefore,
P{D 2 1 £>,) = 3/19
Hence, the required probability is
P{D { and£> 2 ) = P{D { ) P{D 2 \D X ) = (4/20)(3/19) = .0316
The tree diagram in Figure 4.17 shows the selection procedure and the final four outcomes
for this experiment along with their probabilities. The joint probability of D x and D 2 is high-
lighted in the tree diagram.
Calculating the joint probability
of two events.
Conditional probability was discussed in Section 4.4. It is obvious from the formula for
joint probability that if we know the probability of an event A and the joint probability of events
A and B, then we can calculate the conditional probability of B given A.
4.8 Intersection of Events and the Multiplication Rule 165
Calculating Conditional Probability If A and B are two events, then,
P(A and B) P(A and B)
P(B\A) = , , and P(A I B) = , '
V 1 ' P(A) y 1 ; P(B)
given that P(A) * and P(B) # 0.
EXAMPLE 4-25
The probability that a randomly selected student from a college is a senior is .20, and the joint „ , , .. ., .... ,
r j j o 'J Calculating the conditional
probability that the student is a computer science major and a senior is .03. Find the condi- probability of an event
tional probability that a student selected at random is a computer science major given that the
student is a senior.
Solution Let us define the following two events:
A = the student selected is a senior
B = the student selected is a computer science major
From the given information,
P(A) = .20 and P(A and B) = .03
Hence,
P(A and B) 03
P(B\A)= ^ — ' = — = .15
V 1 ; P(A) .20
Thus, the (conditional) probability is .15 that a student selected at random is a computer
science major given that he or she is a senior. H
Multiplication Rule for Independent Events
The foregoing discussion of the multiplication rule was based on the assumption that the two
events are dependent. Now suppose that events A and B are independent. Then,
P(A) = P(A | B) and P(B) = P(B | A)
By substituting P(B) for P(B | A) into the formula for the joint probability of A and B, we
obtain
P(A and B) = P(A) P(B)
Multiplication Rule to Calculate the Probability of Independent Events The probability of the
intersection of two independent events A and B is
P(A and B) = P(A) P(B)
of two independent events.
■ EXAMPLE 4-26
An office building has two fire detectors. The probability is .02 that any fire detector of this
type will fail to go off during a fire. Find the probability that both of these fire detectors will Calculating foe joint probability
fail to go off in case of a fire.
Solution In this example, the two fire detectors are independent because whether or not one
fire detector goes off during a fire has no effect on the second fire detector. We define the fol-
lowing two events:
A = the first fire detector fails to go off during a fire
B = the second fire detector fails to go off during a fire
166 Chapter 4 Probability
Then, the joint probability of A and B is
P(A and B) = P(A) P(B) = (.02)(.02) = .0004 ■
The multiplication rule can be extended to calculate the joint probability of more than two events.
Example 4-27 illustrates such a case for independent events.
Calculating the joint probability
to three patients
oj three events. r
I EXAMPLE 4-27
The probability that a patient is allergic to penicillin is .20. Suppose this drug is administered
(a) Find the probability that all three of them are allergic to it.
(b) Find the probability that at least one of them is not allergic to it.
Solution
(a) Let A, B, and C denote the events that the first, second, and third patients, respec-
tively, are allergic to penicillin. We are to find the joint probability of A, B, and C.
All three events are independent because whether or not one patient is allergic does
not depend on whether or not any of the other patients is allergic. Hence,
P(A and B and C) = P(A) P(B) P(C) = (.20)(.20)(.20) = .008
The tree diagram in Figure 4.18 shows all the outcomes for this experiment. Events
A, B, and C are the complementary events of A, B, and C, respectively. They repre-
sent the events that the patients are not allergic to penicillin. Note that the intersec-
tion of events A, B, and C is written as ABC in the tree diagram.
First patient
Third patient
Final outcomes
P(ABC) = .008
P(ABC) = .032
P(ABC) = .032
P(ABC) = .128
P(ABC) = .032
P(ABC) = .128
P(ABC) = .128
P(ABC) = .512
Figure 4.18 Tree diagram for joint probabilities.
Going "0 for July," as former infielder Bob Aspromonte once put it, is enough to make a baseball
player toss out his lucky bat or start seriously searching for flaws in his hitting technique. But the
culprit is usually just simple mathematics.
Statistician Harry Roberts of the University of Chicago's Graduate School of Business studied
the records of major-league baseball players and found that a batter is no more likely to hit worse
when he is in a slump than when he is in a hot streak. The occurrences of hits followed the same
pattern as purely random events such as pulling marbles out of a hat. If there were one white
marble and three black ones in the hat, for example, then a white marble would come out about
one quarter of the time— a .250 average. In the same way, a player who hits .250 will in the long
run get a hit every four times at bat.
But that doesn't mean the player will hit the ball exactly every fourth time he comes to the
plate— just as it's unlikely that the white marble will come out exactly every fourth time.
Even a batter who goes hitless 10 times in a row might safely be able to pin the blame on
statistical fluctuations. The odds of pulling a black marble out of a hat 10 times in a row are about
6 percent— not a frequent occurrence, but not impossible, either. Only in the long run do these
statistical fluctuations even out.
As mentioned in the above excerpt from U.S. News & World Report, if we assume a player hits .250
in the long run, the probability that this player does not hit during a specific trip to the plate is .75. Hence,
we can calculate the probability that he goes hitless 10 times in a row as follows:
P(hitless 1 times in a row) = (.75)(.75) ■ ■ ■ (.75) ten times
= (.75) 10 = .0563
Note that each trip to the plate is independent, and the probability that a player goes hitless 10 times in
a row is given by the intersection of 10 hitless trips. This probability has been rounded off to "about 6°/o"
in this illustration.
(b) Let us define the following events:
G = all three patients are allergic
H = at least one patient is not allergic
Events G and H are two complementary events. Event G consists of the intersection
of events A, B, and C. Hence, from part (a),
P(G) = P(A and B and C) = .008
Therefore, using the complementary event rule, we obtain
P(H) = 1 - P(G) = 1 - .008 = .992 ■
Case Study 4-2 calculates the probability of a hitless streak in baseball by using the mul-
tiplication rule.
Joint Probability of Mutually Exclusive Events
We know from an earlier discussion that two mutually exclusive events cannot happen together.
Consequently, their joint probability is zero.
Joint Probability of Mutually Exclusive Events The joint probability of two mutually exclusive
events is always zero. If A and B are two mutually exclusive events, then,
P(A and B) =
BASEBALL
PLAYERS
HAVE
"SLUMPS"
AND
"STREAKS"
Source: U.S. News & World Report, July
1 1, 1988, p. 46. Copyright © 1988, by
U.S. News & World Report, Inc. Excerpts
reprinted with permission.
168 Chapter 4 Probability
Illustrating the probability of
two mutually exclusive events.
■ EXAMPLE 4-28
Consider the following two events for an application filed by a person to obtain a car loan:
A = event that the loan application is approved
R = event that the loan application is rejected
What is the joint probability of A and R?
Solution The two events A and R are mutually exclusive. Either the loan application will
be approved or it will be rejected. Hence,
P{A and R) = ■
EXERCISES
CONCEPTS AND PROCEDURES
4.66 Explain the meaning of the intersection of two events. Give one example.
4.67 What is meant by the joint probability of two or more events? Give one example.
4.68 How is the multiplication rule of probability for two dependent events different from the rule for two
independent events?
4.69 What is the joint probability of two mutually exclusive events? Give one example.
4.70 Find the joint probability of A and B for the following.
a. P(A) = .40 and P(B | A) = .25
b. P(B) = .65 and P(A j B) = .36
4.71 Find the joint probability of A and B for the following.
a. P(B) = .59 and P(A | B) = .77
b. P(A) = .28 and P(B j A) = .35
4.72 Given that A and B are two independent events, find their joint probability for the following.
a. P(A) = .61 and P(B) = .27
b. P(A) = .39 and P(B) = .63
4.73 Given that A and B are two independent events, find their joint probability for the following.
a. P(A) = .20 and P(B) = .76
b. P(A) = .57 and P(B) = .32
4.74 Given that A, B, and C are three independent events, find their joint probability for the following.
a. P(A) = .20, P(B) = .46, and P(C) = .25
b. P(A) = .44, P(B) = .27, and P(C) = .43
4.75 Given that A, B, and C are three independent events, find their joint probability for the following.
a. P(A) = .49, P(B) = .67, and P(C) = .75
b. P(A) = .71, P(B) = .34, and P(C) = .45
4.76 Given that P(A) = .30 and P(A and B) = .24, find P(B \ A).
4.77 Given that P(B) = .65 and P(A and B) = .45, find P(A \ B).
4.78 Given that P(A | B) = .40 and P(A and B) = .36, find P{B).
4.79 Given that P(B \ A) = .80 and P(A and B) = .58, find P(A).
■ APPLICATIONS
4.80 In a sample survey, 1 800 senior citizens were asked whether or not they have ever been victimized
by a dishonest telemarketer. The following table gives the responses by age group (in years).
Have Been
Victimized
Have Never
Been Victimized
60-69 (A)
106
698
Age
70-79 (B)
145
447
80 or over (C)
61
343
4.8 Intersection of Events and the Multiplication Rule 169
a. Suppose one person is randomly selected from these senior citizens. Find the following probabilities.
i. P(have been victimized and C)
ii. P(have never been victimized and A)
b. Find P(B and C). Is this probability zero? Explain why or why not.
4.81 The following table gives a two-way classification of all basketball players at a state university who
began their college careers between 2001 and 2005, based on gender and whether or not they graduated.
Graduated
Did Not Graduate
Male
126
55
Female
133
32
a. If one of these players is selected at random, find the following probabilities.
i. /'(female and graduated)
ii. /'(male and did not graduate)
b. Find P(graduated and did not graduate). Is this probability zero? If yes, why?
4.82 Five hundred employees were selected from a city's large private companies and asked whether or
not they have any retirement benefits provided by their companies. Based on this information, the follow-
ing two-way classification table was prepared.
Have Retirement Benefits
Yes No
Men
225 75
Women
150 50
a. Suppose one employee is selected at random from these 500 employees. Find the following prob-
abilities.
i. Probability of the intersection of events "woman" and "yes"
ii. Probability of the intersection of events "no" and "man"
b. Mention what other joint probabilities you can calculate for this table and then find them. You
may draw a tree diagram to find these probabilities.
4.83 Two thousand randomly selected adults were asked whether or not they have ever shopped on the
Internet. The following table gives a two-way classification of the responses obtained.
Have Shopped
Have Never Shopped
Male
500
700
Female
300
500
a. Suppose one adult is selected at random from these 2000 adults. Find the following probabilities.
i. /'(has never shopped on the Internet and is a male)
ii. P(has shopped on the Internet and is a female)
b. Mention what other joint probabilities you can calculate for this table and then find those. You
may draw a tree diagram to find these probabilities.
4.84 A consumer agency randomly selected 1700 flights for two major airlines, A and B. The following
table gives the two-way classification of these flights based on airline and arrival time. Note that "less than
30 minutes late" includes flights that arrived early or on time.
Less Than 30
30 Minutes to
More Than
Minutes Late
1 Hour Late
1 Hour Late
Airline A
429
390
92
Airline B
393
316
80
a. Suppose one flight is selected at random from these 1700 flights. Find the following probabilities.
i. /'(more than 1 hour late and airline A)
ii. /'(airline B and less than 30 minutes late)
b. Find the joint probability of events "30 minutes to 1 hour late" and "more than 1 hour late." Is
this probability zero? Explain why or why not.
1 70 Chapter 4 Probability
4.85 Two thousand randomly selected adults were asked if they think they are financially better off than
their parents. The following table gives the two-way classification of the responses based on the educa-
tion levels of the persons included in the survey and whether they are financially better off, the same as,
or worse off than their parents.
Less Than
High
More Than
High School
School
High School
Better off
140
450
420
Same as
60
250
110
Worse off
200
300
70
a. Suppose one adult is selected at random from these 2000 adults. Find the following probabilities.
i. P(better off and high school)
ii. P(more than high school and worse off)
b. Find the joint probability of the events "worse off" and "better off." Is this probability zero? Ex-
plain why or why not.
4.86 In a statistics class of 42 students, 28 have volunteered for community service in the past. If two stu-
dents are selected at random from this class, what is the probability that both of them have volunteered
for community service in the past? Draw a tree diagram for this problem.
4.87 In a political science class of 35 students, 21 favor abolishing the electoral college and thus electing
the President of the United States by popular vote. If two students are selected at random from this class,
what is the probability that both of them favor abolition of the electoral college? Draw a tree diagram for
this problem.
4.88 A company is to hire two new employees. They have prepared a final list of eight candidates, all of
whom are equally qualified. Of these eight candidates, five are women. If the company decides to select
two persons randomly from these eight candidates, what is the probability that both of them are women?
Draw a tree diagram for this problem.
4.89 In a group of 10 persons, 4 have a type A personality and 6 have a type B personality. If two persons
are selected at random from this group, what is the probability that the first of them has a type A per-
sonality and the second has a type B personality? Draw a tree diagram for this problem.
4.90 The probability is .80 that a senior from a large college in New York State has never gone to Florida
for spring break. If two college seniors are selected at random from this college, what is the probability
that the first has never gone to Florida for spring break and the second has? Draw a tree diagram for this
problem.
4.91 The probability that a student graduating from Suburban State University has student loans to pay
off after graduation is .60. If two students are randomly selected from this university, what is the proba-
bility that neither of them has student loans to pay off after graduation?
4.92 A contractor has submitted bids for two state construction projects. The probability of winning each
contract is .25, and it is the same for both contracts.
a. What is the probability that he will win both contracts?
b. What is the probability that he will win neither contract?
Draw a tree diagram for this problem.
4.93 Five percent of all items sold by a mail-order company are returned by customers for a refund. Find
the probability that, of two items sold during a given hour by this company,
a. both will be returned for a refund
b. neither will be returned for a refund
Draw a tree diagram for this problem.
4.94 The probability that any given person is allergic to a certain drug is .03. What is the probability that
none of three randomly selected persons is allergic to this drug? Assume that all three persons are inde-
pendent.
4.95 The probability that a farmer is in debt is .80. What is the probability that three randomly selected
farmers are all in debt? Assume independence of events.
4.96 The probability that a student graduating from Suburban State University has student loans to pay
off after graduation is .60. The probability that a student graduating from this university has student loans
4.9 Union of Events and the Addition Rule
to pay off after graduation and is a male is .24. Find the conditional probability that a randomly selected
student from this university is a male given that this student has student loans to pay off after graduation.
4.97 The probability that an employee at a company is a female is .36. The probability that an employee
is a female and married is .19. Find the conditional probability that a randomly selected employee from
this company is married given that she is a female.
4.98 A telephone poll conducted of 1000 adult Americans for the Washington Post in March 2009 asked
about current events in the United States. Suppose that of the 1000 respondents, 629 stated that they were
cutting back on their daily spending. Suppose that 322 of the 629 people who stated that they were cut-
ting back on their daily spending said that they were cutting back "somewhat" and 97 stated that they were
cutting back "somewhat" and delaying the purchase of a new car by at least 6 months. If one of the 629
people who are cutting back on their spending is selected at random, what is the probability that he/she
is delaying the purchase of a new car by at least 6 months given that he/she is cutting back on spending
"somewhat?"
4.99 Suppose that 20% of all adults in a small town live alone, and 8% of the adults live alone and have
at least one pet. What is the probability that a randomly selected adult from this town has at least one pet
given that this adult lives alone?
4.9 Union of Events and the Addition Rule
This section discusses the union of events and the addition rule that is applied to compute the
probability of the union of events.
4.9.1 Union of Events
The union of two events A and B includes all outcomes that are either in A or in B or in both
A and B.
Definition
Union of Events Let A and B be two events defined in a sample space. The union of events A
and B is the collection of all outcomes that belong either to A or to B or to both A and B and is
denoted by
A orB
The union of events A and B is also denoted by AUB. Example 4-29 illustrates the union
of events A and B.
■ EXAMPLE 4-29
A senior citizens center has 300 members. Of them, 140 are male, 210 take at least one
medicine on a permanent basis, and 95 are male and take at least one medicine on a per-
manent basis. Describe the union of the events "male" and "take at least one medicine on a
permanent basis."
Solution Let us define the following events:
M = a senior citizen is a male
F = a senior citizen is a female
A = a senior citizen takes at least one medicine
B = a senior citizen does not take any medicine
The union of the events "male" and "take at least one medicine" includes those senior citizens
who are either male or take at least one medicine or both. The number of such senior citizens is
140 + 210 - 95 = 255
Illustrating the union
of two events.
172 Chapter 4 Probability
Why did we subtract 95 from the sum of 140 and 210? The reason is that 95 senior citizens
(which represent the intersection of events M and A) are common to both events M and A
and, hence, are counted twice. To avoid double counting, we subtracted 95 from the sum of
the other two numbers. We can observe this double counting from Table 4.8, which is con-
structed using the given information. The sum of the numbers in the three shaded cells gives
the number of senior citizens who are either male or take at least one medicine or both. How-
ever, if we add the totals of the row labeled M and the column labeled A, we count 95 twice.
Table 4.8
A
B
Total
M
m
140
F
m
45
160
Total
210
90
300
> Counted twice
Figure 4.19 shows the diagram for the union of the events "male" and "take at least one med-
icine on a permanent basis."
Figure 4.19 Union of events
M and A.
Area shaded in red gives the
union of events M and A, and
includes 255 senior citizens I
4.9.2 Addition Rule
The method used to calculate the probability of the union of events is called the addition rule.
It is defined as follows.
Addition Rule to Find the Probability of Union of Events The probability of the union of two
events A and B is
P(A or B) = P(A) + P(B) - P(A and B)
Thus, to calculate the probability of the union of two events A and B, we add their marginal
probabilities and subtract their joint probability from this sum. We must subtract the joint proba-
bility of A and B from the sum of their marginal probabilities to avoid double counting because of
common outcomes in A and B. This is the case where events A and B are not mutually exclusive.
■ EXAMPLE 4-30
A university president proposed that all students must take a course in ethics as a requirement
for graduation. Three hundred faculty members and students from this university were asked
about their opinions on this issue. Table 4.9 gives a two-way classification of the responses of
these faculty members and students.
Table 4.9 Two-Way Classification of Responses
Favor
Oppose
Neutral
Total
Faculty
45
15
10
70
Student
90
110
30
230
Fotal
135
125
40
300
Calculating the probability
of the union of two events:
two-way table.
4.9 Union of Events and the Addition Rule
Find the probability that one person selected at random from these 300 persons is a faculty
member or is in favor of this proposal.
Solution Let us define the following events:
A = the person selected is a faculty member
B = the person selected is in favor of the proposal
From the information given in Table 4.9,
P(A) = 70/300 = .2333
P(B) = 135/300 = .4500
P{A and B) = P{A) P{B\A) = (70/300)(45/70) = .1500
Using the addition rule, we obtain
P(A or 5) = P(A) + P(B) - P(A andfi) = .2333 + .4500 - .1500 = .5333
Thus, the probability that a randomly selected person from these 300 persons is a faculty mem-
ber or is in favor of this proposal is .5333.
The probability in this example can also be calculated without using the addition rule. The total
number of persons in Table 4.9 who are either faculty members or in favor of this proposal is
45 + 15 + 10 + 90 = 160
Hence, the required probability is
P(A or 5) = 160/300 = .5333 ■
■ EXAMPLE 4-31
In a group of 2500 persons, 1400 are female, 600 are vegetarian, and 400 are female and veg-
etarian. What is the probability that a randomly selected person from this group is a male or
vegetarian?
Solution Let us define the following events:
F = the randomly selected person is a female
M = the randomly selected person is a male
V = the randomly selected person is a vegetarian
N = the randomly selected person is a non-vegetarian
From the given information, we know that of the group, 1400 are female, 600 are vegetarian, and
400 are female and vegetarian. Hence, 1100 are male, 1900 are nonvegetarian, and 200 are male
and vegetarian. We are to find the probability P(M or V). This probability is obtained as follows:
P(Mor V) = P(M) + P(V) - P(M and V)
1100 600 _ 200
~ 2500 2500 2500
= .44 + .24 - .08 = .60
Actually, using the given information, we can prepare Table 4. 10 for this example. In the table,
the numbers in the shaded cells are given to us. The remaining numbers are calculated by do-
ing some arithmetic manipulations.
Table 4.10
Two- Way Classification Table
Vegetarian (V)
Nonvegetarian (AO
Total
Female (F)
[400]
1000
1 1400 1
Male (M)
200
900
1100
Total
[Ml
1900
2500
Calculating the probability of
the union of two events.
174 Chapter 4 Probability
Using Table 4.10, we find the required probability:
P(M or V) = P(M) + P(V) - P(M and V)
1100 600 200
+ = .44 + .24 - .08 = .60
2500 2500 2500
Addition Rule for Mutually Exclusive Events
We know from an earlier discussion that the joint probability of two mutually exclusive events
is zero. When A and B are mutually exclusive events, the term P(A and B) in the addition rule
becomes zero and is dropped from the formula. Thus, the probability of the union of two mu-
tually exclusive events is given by the sum of their marginal probabilities.
Addition Rule to Find the Probability of the Union of Mutually Exclusive Events The probability
of the union of two mutually exclusive events A and B is
P(A or B) = P(A) + P(B)
Calculating the probability of
the union of two mutually
exclusive events: two-way table.
■ EXAMPLE 4-32
A university president proposed that all students must take a course in ethics as a requirement
for graduation. Three hundred faculty members and students from this university were asked
about their opinion on this issue. The following table, reproduced from Table 4.9 in Example
4-30, gives a two-way classification of the responses of these faculty members and students.
Favor
Oppose
Neutral
Total
Faculty
45
15
10
70
Student
90
110
30
230
Total
135
125
40
300
What is the probability that a randomly selected person from these 300 faculty members and
students is in favor of the proposal or is neutral?
Solution Let us define the following events:
F = the person selected is in favor of the proposal
N = the person selected is neutral
As shown in Figure 4.20, events F and N are mutually exclusive because a person selected
can be either in favor or neutral but not both.
Figure 4.20 Venn diagram of mutually
exclusive events.
From the given information,
Hence,
P{F) = 135/300 = .4500
P(N) = 40/300 = .1333
P(F or TV) = P(F) + P(N) = .4500 + .1333 = .5833
4.9 Union of Events and the Addition Rule 175
The addition rule formula can easily be extended to apply to more than two events. The fol-
lowing example illustrates this.
EXAMPLE 4-33
Consider the experiment of rolling a die twice. Find the probability that the sum of the num-
bers obtained on two rolls is 5, 7, or 10.
Solution The experiment of rolling a die twice has a total of 36 outcomes, which are
listed in Table 4.11. Assuming that the die is balanced, these 36 outcomes are equally
likely.
Calculating the probability of
the union of three mutually ex-
clusive events.
Table 4.11 Two Rolls of a Die
1
2
Second Roll of the Die
3 4
5
6
1
(1,1)
(1,2)
^TM))
(1^5^-
---TX§)
2
(2,1)
(2^-
^^(23)^
-""(16)
First
Roll of
3
(3J^
^02}^
-^(33)^-
-^04)^^
-"05)
(3,6)
the Die
4
5
(54}^
^""""(53)
-""(^6)
6
((<vL^
-"""(^2)
(6,3)
(6,6)
The events that give the sum of two numbers equal to 5 or 7 or 10 are shaded in the table.
As we can observe, the three events "the sum is 5," "the sum is 7," and "the sum is 10" are
mutually exclusive. Four outcomes give a sum of 5, six give a sum of 7, and three outcomes
give a sum of 10. Thus,
P(sum is 5 or 7 or 10) = P(sum is 5) + P(sum is 7) + P(sum is 10)
= 4/36 + 6/36 + 3/36 = 13/36 = .3611 ■
■ EXAMPLE 4-34
The probability that a person is in favor of genetic engineering is .55 and that a person is
against it is .45. Two persons are randomly selected, and it is observed whether they favor or Calculating the probability
oppose genetic engineering. °f the union °f three mutua ^
exclusive events.
(a) Draw a tree diagram for this experiment.
(b) Find the probability that at least one of the two persons favors genetic engineering.
Solution
(a) Let
F = a person is in favor of genetic engineering
A = a person is against genetic engineering
This experiment has four outcomes: both persons are in favor (FF), the first person is
in favor and the second is against (FA), the first person is against and the second is in
favor (AF), and both persons are against genetic engineering (AA). The tree diagram
in Figure 4.21 shows these four outcomes and their probabilities.
176 Chapter 4 Probability
First person
Second person
Final outcomes and
their probabilities
P{FF) = (.55) (.55) = .3025
P(FA) = (.55) (.45) = .2475
P(AF) = (.45) (.55) = .2475
P(AA) = (.45) (.45) = .2025
Figure 4.21 Tree diagram.
(b) The probability that at least one person favors genetic engineering is given by the
union of events FF, FA, and AF. These three outcomes are mutually exclusive. Hence,
P(at least one person favors) = P(FF or FA or AF)
= P(FF) + P(FA) + P(AF)
= .3025 + .2475 + .2475 = .7975 ■
I
EXERCISES
CONCEPTS AND PROCEDURES
4.100 Explain the meaning of the union of two events. Give one example.
4.101 How is the addition rule of probability for two mutually exclusive events different from the rule for
two mutually nonexclusive events?
4.102 Consider the following addition rule to find the probability of the union of two events A and B:
P(A or B) = P(A) + P(B) - P(A and B)
When and why is the term P(A and B) subtracted from the sum of P(A) and P{B)1 Give one example where
you might use this formula.
4.103 When is the following addition rule used to find the probability of the union of two events A and fi?
P(A or B) = P(A) + P(B)
Give one example where you might use this formula.
4.104 Find P(A or B) for the following.
a. P(A) = .58, P(B) = .66, and P(A and B) = .57
b. P(A) = .72, P(B) = .42, and P(A and B) = .39
4.105 Find P(A or B) for the following.
a. P(A) = .18, P(B) = .49, and P(A and S) = .11
b. P(A) = .73, P(B) = .71, and P(A and B) = .68
4.106 Given that A and B are two mutually exclusive events, find P(A or B) for the following.
a. P(A) = Al and P(B) = .32
b. P(A) = .16 and P(B) = .59
4.107 Given that A and B are two mutually exclusive events, find P(A or B) for the following,
a. P(A) = .25 and P(B) = .27
b. P(A) = .58 and P(B)
.09
4.9 Union of Events and the Addition Rule
177
■ APPLICATIONS
4.108 In a sample survey, 1800 senior citizens were asked whether or not they have ever been victimized
by a dishonest telemarketer. The following table gives the responses by age group.
Have Been
Victimized
Have Never
Been Victimized
60-69 (A)
106
698
Age
70-79 (B)
145
447
80 or over (C)
61
343
Suppose one person is randomly selected from these senior citizens. Find the following probabilities.
a. P(have been victimized or B)
b. P(have never been victimized or C)
4.109 The following table gives a two-way classification of all basketball players at a state university
who began their college careers between 2001 and 2005, based on gender and whether or not they
graduated.
Graduated
Did Not Graduate
Male
126
55
Female
133
32
If one of these players is selected at random, find the following probabilities.
a. P(female or did not graduate)
b. P(graduated or male)
4.110 Five hundred employees were selected from a city's large private companies, and they were asked
whether or not they have any retirement benefits provided by their companies. Based on this information,
the following two-way classification table was prepared.
Have Retirement Benefits
Yes No
Men
225 75
Women
150 50
Suppose one employee is selected at random from these 500 employees. Find the following probabilities.
a. The probability of the union of events "woman" and "yes"
b. The probability of the union of events "no" and "man"
4.111 Two thousand randomly selected adults were asked whether or not they have ever shopped on the
Internet. The following table gives a two-way classification of the responses.
Have Shopped
Have Never Shopped
Male
500
700
Female
300
500
Suppose one adult is selected at random from these 2000 adults. Find the following probabilities.
a. .P(has never shopped on the Internet or is a female)
b. P(is a male or has shopped on the Internet)
c. />(has shopped on the Internet or has never shopped on the Internet)
4.112 A consumer agency randomly selected 1700 flights for two major airlines, A and B. The follow-
ing table gives the two-way classification of these flights based on airline and arrival time. Note that "less
than 30 minutes late" includes flights that arrived early or on time.
178 Chapter 4 Probability
Less Than 30
30 Minutes to
More Than
Minutes Late
1 Hour Late
1 Hour Late
Airline A
429
390
92
Airline B
393
316
80
If one flight is selected at random from these 1700 flights, find the following probabilities.
a. /"(more than 1 hour late or airline A)
b. P(airline B or less than 30 minutes late)
c. /"(airline A or airline B)
4.113 Two thousand randomly selected adults were asked if they think they are financially better off than
their parents. The following table gives the two-way classification of the responses based on the educa-
tion levels of the persons included in the survey and whether they are financially better off, the same as,
or worse off than their parents.
Less Than
High
More Than
High School
School
High School
Better off
140
450
420
Same as
60
250
110
Worse off
200
300
70
Suppose one adult is selected at random from these 2000 adults. Find the following probabilities.
a. /"(better off or high school)
b. /"(more than high school or worse off)
c. /"(better off or worse off)
4.114 There is an area of free (but illegal) parking near an inner-city sports arena. The probability that a
car parked in this area will be ticketed by police is .35, that the car will be vandalized is .15, and that it
will be ticketed and vandalized is .10. Find the probability that a car parked in this area will be ticketed
or vandalized.
4.115 The probability that a family owns a washing machine is .68, that it owns a DVD player is .81, and
that it owns both a washing machine and a DVD player is .58. What is the probability that a randomly se-
lected family owns a washing machine or a DVD player?
4.116 Jason and Lisa are planning an outdoor reception following their wedding. They estimate that the
probability of bad weather is .25, that of a disruptive incident (a fight breaks out, the limousine is late, etc.)
is .15, and that bad weather and a disruptive incident will occur is .08. Assuming these estimates are cor-
rect, find the probability that their reception will suffer bad weather or a disruptive incident.
4.117 The probability that a randomly selected elementary or secondary school teacher from a city is a
female is .68, holds a second job is .38, and is a female and holds a second job is .29. Find the probabil-
ity that an elementary or secondary school teacher selected at random from this city is a female or holds a
second job.
4.118 According to the U.S. Census Bureau's most recent data on the marital status of the 238 million
Americans aged 15 years and older, 123.7 million are currently married and 71.5 million have never
been married. If one person from these 238 million persons is selected at random, find the probability
that this person is currently married or has never been married. Explain why this probability is not equal
to 1.0.
4.119 According to a survey of 2000 home owners, 800 of them own homes with three bedrooms, and
600 of them own homes with four bedrooms. If one home owner is selected at random from these 2000
home owners, find the probability that this home owner owns a house that has three or four bedrooms.
Explain why this probability is not equal to 1.0.
4.120 The probability of a student getting an A grade in an economics class is .24 and that of getting a
B grade is .28. What is the probability that a randomly selected student from this class will get an A or a
B in this class? Explain why this probability is not equal to 1.0.
4.121 Twenty percent of a town's voters favor letting a major discount store move into their neighbor-
hood, 63% are against it, and 17% are indifferent. What is the probability that a randomly selected voter
from this town will either be against it or be indifferent? Explain why this probability is not equal to 1.0.
Uses and Misuses 179
4.122 The probability that a corporation makes charitable contributions is .72. Two corporations are
selected at random, and it is noted whether or not they make charitable contributions.
a. Draw a tree diagram for this experiment.
b. Find the probability that at most one corporation makes charitable contributions.
4.123 The probability that an open-heart operation is successful is .84. What is the probability that in two ran-
domly selected open-heart operations at least one will be successful? Draw a tree diagram for this experiment.
USES AND MISUSES...
1. STATISTICS VERSUS PROBABILITY
At this point, you may think that probability and statistics are basically
the same things. They both use the term mean, they both report results
in terms of percentages, and so on. Do not be fooled: Although they
share many of the same mathematical tools, probability and statistics
are very different sciences. The first three chapters of the text were very
careful to specify whether a particular set of data was a population or
a sample. This is because statistics takes a sample of data and, based
upon the properties of that sample— mean, median, mode, standard
deviation— attempts to say something about a population. Probability
does exactly the opposite: In probability, we know the properties of the
population based on the sample space and the probability distribution,
and we want to make statements about a sample from the population.
Here's an example viewed from a statistical and a probabilistic
point of view. A sequence of outcomes from 10 independent coin
tosses is {H, T, H, T, H, T, T, H, T, T}. A statistician will ask the ques-
tion: Based on the observed 4 heads and 6 tails, what combination
of heads and tails would he or she expect from 100 or 1000 tosses,
and how certain would he or she be of that answer? Someone us-
ing probability will ask: If the coin toss was fair (the probability of
the event that a single coin toss be a head or tail is .5), what is the
probability that the compound event of four heads and six tails will
occur? These are substantially different questions.
The distinction between a statistical approach and a probabilis-
tic approach to a problem can be surprising. Imagine that you must
determine the average life of an automotive part. One approach would
be to take a sample of parts, test each of them until they fail to work,
and then perform some calculations regarding the distribution of fail-
ures. However, if this particular part has outliers with long life spans
(several years), you are going to be spending a lot of time in the lab-
oratory. An approach using probabilistic techniques could develop a
hypothetical life span based on the physical properties of the part, the
conditions of its use, and the manufacturing characteristics. Then you
can use your experimental results over a relatively short period of
time— including data on those parts that did not fail— to adjust your
prior understanding of what makes the part fail, saving yourself a lot
of time.
2. ODDS AND PROBABILITY
One of the first things we learn in probability is that the sum of the
probabilities of all outcomes for an experiment must equal 1.0. We
also learn about the probabilities that are developed from relative
frequencies and about subjective probabilities. In the latter case, many
of the probabilities involve personal opinions— one hopes, those of
experts in the field. Still, both scenarios (probabilities obtained from
relative frequencies and subjective probabilities) require that all
Team
Odds
Team
Odds
Arizona Diamondbacks
1:12
Milwaukee Brewers
1:28
Atlanta Braves
1:60
Minnesota Twins
1:16
Baltimore Orioles
1:80
New York Mets
1:16
Boston Red Sox
2:13
New York Yankees
1:9
Chicago Cubs
2:15
Oakland Athletics
1:40
Chicago White Sox
1:30
Philadelphia Phillies
1:12
Cincinnati Reds
1:80
Pittsburgh Pirates
1:150
Cleveland Indians
1:25
San Diego Padres
1:100
Colorado Rockies
1:38
San Francisco Giants
1:60
Detroit Tigers
1:30
Seattle Mariners
1:80
Florida Marlins
1:20
St. Louis Cardinals
1:18
Houston Astros
1:40
Tampa Bay Devil Rays
1:10
Kansas City Royals
1:125
Texas Rangers
1:75
Los Angeles Angels
1:8
Toronto Blue Jays
1:35
Los Angeles Dodgers
1:8
Washington Nationals
1:125
180 Chapter 4 Probability
probabilities must be nonnegative and the sum of the probabilities
of all outcomes must equal 1.0.
Although probabilities and probability models are all around us—
in weather, medicine, financial markets, and so forth— they are most
obvious in the world of gaming and gambling. Sports betting agencies
publish odds of various teams winning specific games or sports titles.
The accompanying table gives the 2009 Opening Day odds for each
Major League Baseball team to win the 2009 World Series. These odds
are obtained from the Web site http://www.vegas.com/gaming/futures/
worldseries.html. (Actually NY Yankees won the 2009 World Series.)
Note that the odds listed in this table are called the odds in fa-
vor of winning the World Series. For example, the Los Angeles Dodgers
have 1 :8 (which is read as 1 to 8) odds to win the 2009 World Se-
ries. If we switch the numbers around, we can state that odds are
8:1 (or 8 to 1) against the Dodgers to win the 2009 World Series.
How do we convert these odds into probabilities? Let us con-
sider the Los Angeles Dodgers. Odds of 1 :8 imply that out of 9
chances, there is 1 chance that the Dodgers will win the 2009 World
Series and 8 chances that the Dodgers will not win the 2009 World Se-
ries. Thus, the probability that the Dodgers will win the 2009 World
Series is - - - = — = .1111, and the probability that Dodgers will not
8 8
win the 2009 World Series is = — = .8888. Similarly, for the
8 + 19
Boston Red Sox, the probability of winning the 2009 World Series is
2 2
— - - = — = .1333, and the probability of not winning the World
Series in 2009 is ^ = jf = - 8667 - ^ e can calculate these prob-
abilities for all teams listed in the table by using this procedure.
Note that here the outcomes that these teams win the 2009
World Series are mutually exclusive events because it is impossible
for two or more teams to win the World Series during the same
year. Hence, if we add the probabilities of winning the 2009 World
Series for all teams, we should obtain a value of 1 .0. However, if
you calculate the probability of winning the 2009 World Series for
each team using the odds given in the table and then add all these
probabilities, the sum will be 1.391166825. So, what happened?
Did these odds makers flunk their statistics and probability courses?
Probably not.
Casinos and odds makers, which are in the business of making
money, are interested in encouraging people to gamble. These prob-
abilities, which seem to violate the basic rules of probability theory,
still obey the primary rule for the casinos, which is that, on average,
a casino is going to make a profit. How does this happen? We will
explore this in the Uses and Misuses section of Chapter 5.
Note: When casinos create odds for sports betting, they recognize that
many people will bet on one of their favorite teams such as the New
York Yankees or the Chicago Cubs. To meet the rule that the sum of
all probabilities is 1.0, the probabilities for the teams more likely to
win would have to be lowered. Lowering a probability corresponds to
lowering the odds. For example, if the odds for the Chicago Cubs were
lowered from 2:15 to 1 :20, the probability for them to win would de-
crease from .1 176 to .0476. If the Cubs remain as one of the favorites
(they currently have the second best odds, after the Boston Red Sox),
many people would bet on them. However, if they win, the casino
would have to pay $20 for every Si bet instead of $7.50 (Si 5.00/2)
for every Si bet. The casinos do not want to do this, and, hence, they
ignore the probability rule in order to make more money. However,
the casinos cannot do this with their traditional games, which are
bound by the standard rules. From a mathematical standpoint, it is
not acceptable to ignore the rule that the probabilities of all final out-
comes for an experiment add up to 1.0
GLossary
Classical probability rule The method of assigning probabilities
to outcomes or events of an experiment with equally likely outcomes.
Complementary events Two events that taken together include all
the outcomes for an experiment but do not contain any common
outcome.
Compound event An event that contains more than one outcome
of an experiment. It is also called a composite event.
Conditional probability The probability of an event subject to the
condition that another event has already occurred.
Dependent events Two events for which the occurrence of one
changes the probability of the occurrence of the other.
Equally likely outcomes Two (or more) outcomes or events that
have the same probability of occurrence.
Event A collection of one or more outcomes of an experiment.
Experiment A process with well-defined outcomes that, when
performed, results in one and only one of the outcomes per repe-
tition.
Impossible event An event that cannot occur.
Independent events Two events for which the occurrence of one
does not change the probability of the occurrence of the other.
Intersection of events The intersection of events is given by the
outcomes that are common to two (or more) events.
Joint probability The probability that two (or more) events occur
together.
Law of Large Numbers If an experiment is repeated again and
again, the probability of an event obtained from the relative fre-
quency approaches the actual or theoretical probability.
Marginal probability The probability of one event or character-
istic without consideration of any other event.
Mutually exclusive events Two or more events that do not contain
any common outcome and, hence, cannot occur together.
Outcome The result of the performance of an experiment.
Probability A numerical measure of the likelihood that a specific
event will occur.
Supplementary Exercises 181
Relative frequency as an approximation of probability Proba-
bility assigned to an event based on the results of an experiment or
based on historical data.
Sample point An outcome of an experiment.
Sample space The collection of all sample points or outcomes of
an experiment.
Simple event An event that contains one and only one outcome of
an experiment. It is also called an elementary event.
Subjective probability The probability assigned to an event based
on the information and judgment of a person.
Sure event An event that is certain to occur.
Tree diagram A diagram in which each outcome of an experiment
is represented by a branch of a tree.
Union of two events Given by the outcomes that belong either to
one or to both events.
Venn diagram A picture that represents a sample space or spe-
cific events.
Supplementary Exercises
4.124 A car rental agency currently has 44 cars available, 28 of which have a GPS navigation system.
One of the 44 cars is selected at random. Find the probability that this car
a. has a GPS navigation system
b. does not have a GPS navigation system
4.125 In a class of 35 students, 13 are seniors, 9 are juniors, 8 are sophomores, and 5 are freshmen. If
one student is selected at random from this class, what is the probability that this student is
a. a junior?
b. a freshman?
4.126 A random sample of 250 juniors majoring in psychology or communication at a large university
is selected. These students are asked whether or not they are happy with their majors. The following table
gives the results of the survey. Assume that none of these 250 students is majoring in both areas.
Happy
Unhappy
Psychology
80
20
Communication
115
35
a. If one student is selected at random from this group, find the probability that this student is
i. happy with the choice of major
ii. a psychology major
iii. a communication major given that the student is happy with the choice of major
iv. unhappy with the choice of major given that the student is a psychology major
v. a psychology major and is happy with that major
vi. a communication major or is unhappy with his or her major
b. Are the events "psychology major" and "happy with major" independent? Are they mutually ex-
clusive? Explain why or why not.
4.127 A random sample of 250 adults was taken, and they were asked whether they prefer watching
sports or opera on television. The following table gives the two-way classification of these adults.
Prefer Watching
Prefer Watching
Sports
Opera
Male
96
24
Female
45
85
a. If one adult is selected at random from this group, find the probability that this adult
i. prefers watching opera
ii. is a male
iii. prefers watching sports given that the adult is a female
iv. is a male given that he prefers watching sports
v. is a female and prefers watching opera
vi. prefers watching sports or is a male
b. Are the events "female" and "prefers watching sports" independent? Are they mutually exclu-
sive? Explain why or why not.
182 Chapter 4 Probability
4.128 A random sample of 80 lawyers was taken, and they were asked if they are in favor of or against
capital punishment. The following table gives the two-way classification of their responses.
Favors Capital
Opposes Capital
Punishment
Punishment
Male
32
24
Female
13
11
a. If one lawyer is randomly selected from this group, find the probability that this lawyer
i. favors capital punishment
ii. is a female
iii. opposes capital punishment given that the lawyer is a female
iv. is a male given that he favors capital punishment
v. is a female and favors capital punishment
vi. opposes capital punishment or is a male
b. Are the events "female" and "opposes capital punishment" independent? Are they mutually ex-
clusive? Explain why or why not.
4.129 A random sample of 400 college students was asked if college athletes should be paid. The follow-
ing table gives a two-way classification of the responses.
Should Be Paid
Should Not Be Paid
Student athlete
90
10
Student nonathlete
210
90
a. If one student is randomly selected from these 400 students, find the probability that this
student
i. is in favor of paying college athletes
ii. favors paying college athletes given that the student selected is a nonathlete
iii. is an athlete and favors paying student athletes
iv. is a nonathlete or is against paying student athletes
b. Are the events "student athlete" and "should be paid" independent? Are they mutually exclu-
sive? Explain why or why not.
4.130 An appliance repair company that makes service calls to customers' homes has found that 5% of
the time there is nothing wrong with the appliance and the problem is due to customer error (appliance
unplugged, controls improperly set, etc.). Two service calls are selected at random, and it is observed
whether or not the problem is due to customer error. Draw a tree diagram. Find the probability that in this
sample of two service calls
a. both problems are due to customer error
b. at least one problem is not due to customer error
4.131 According to the May 2009 issue of U.S. News and World Report, 85.1% of the students who grad-
uated with an MBA degree in 2008 from the University of Virginia's Darden School of Business had job
offers before the graduation date. Suppose that this percentage is true for the top 50 MBA programs in
the list of 426 MBA programs analyzed in this issue of the U.S. News and World Report. Suppose that
two 2008 MBA graduates are selected at random from these top 50 MBA programs and asked if they had
job offers before the graduation date. Draw a tree diagram for this problem. Find the probability that in
this sample of two graduates
a. both had job offers before the graduation date
b. at most one had job offer before the graduation date
4.132 Refer to Exercise 4.124. Two cars are selected at random from these 44 cars. Find the probability
that both of these cars have GPS navigation systems.
4.133 Refer to Exercise 4.125. Two students are selected at random from this class of 35 students. Find
the probability that the first student selected is a junior and the second is a sophomore.
4.134 A company has installed a generator to back up the power in case there is a power failure. The
probability that there will be a power failure during a snowstorm is .30. The probability that the genera-
tor will stop working during a snowstorm is .09. What is the probability that during a snowstorm the com-
pany will lose both sources of power? Note that the two sources of power are independent.
Supplementary Exercises 183
4.135 Terry & Sons makes bearings for autos. The production system involves two independent process-
ing machines so that each bearing passes through these two processes. The probability that the first pro-
cessing machine is not working properly at any time is .08, and the probability that the second machine
is not working properly at any time is .06. Find the probability that both machines will not be working
properly at any given time.
Advanced Exercises
4.136 A player plays a roulette game in a casino by betting on a single number each time. Because the
wheel has 38 numbers, the probability that the player will win in a single play is 1/38. Note that each
play of the game is independent of all previous plays.
a. Find the probability that the player will win for the first time on the 10th play.
b. Find the probability that it takes the player more than 50 plays to win for the first time.
c. The gambler claims that because he has 1 chance in 38 of winning each time he plays, he is cer-
tain to win at least once if he plays 38 times. Does this sound reasonable to you? Find the prob-
ability that he will win at least once in 38 plays.
4.137 A certain state's auto license plates have three letters of the alphabet followed by a three-digit
number.
a. How many different license plates are possible if all three-letter sequences are permitted and any
number from 000 to 999 is allowed?
b. Arnold witnessed a hit-and-run accident. He knows that the first letter on the license plate of the
offender's car was a B, that the second letter was an O or a Q, and that the last number was a
5. How many of this state's license plates fit this description?
4.138 The median life of Brand LT5 batteries is 100 hours. What is the probability that in a set of three
such batteries, exactly two will last longer than 100 hours?
4.139 Powerball is a game of chance that has generated intense interest because of its large jackpots.
To play this game, a player selects five different numbers from 1 through 59, and then picks a Powerball
number from 1 through 39. The lottery organization randomly draws 5 different white balls from 59 balls
numbered 1 through 59, and then randomly picks a Powerball number from 1 through 39. Note that it is
possible for the Powerball number to be the same as one of the first five numbers.
a. If the player's first five numbers match the numbers on the five white balls drawn by the lottery
organization and the player's Powerball number matches the Powerball number drawn by the
lottery organization, the player wins the jackpot. Find the probability that a player who buys
one ticket will win the jackpot. (Note that the order in which the five white balls are drawn is
unimportant.)
b. If the player's first five numbers match the numbers on the five white balls drawn by the lottery
organization, the player wins about $200,000. Find the probability that a player who buys one
ticket will win this prize.
4.140 A trimotor plane has three engines — a central engine and an engine on each wing. The plane will
crash only if the central engine fails and at least one of the two wing engines fails. The probability of fail-
ure during any given flight is .005 for the central engine and .008 for each of the wing engines. Assuming
that the three engines operate independently, what is the probability that the plane will crash during a flight?
4.141 A box contains 10 red marbles and 10 green marbles.
a. Sampling at random from the box five times with replacement, you have drawn a red marble all
five times. What is the probability of drawing a red marble the sixth time?
b. Sampling at random from the box five times without replacement, you have drawn a red mar-
ble all five times. Without replacing any of the marbles, what is the probability of drawing a red
marble the sixth time?
c. You have tossed a fair coin five times and have obtained heads all five times. A friend argues
that according to the law of averages, a tail is due to occur and, hence, the probability of obtain-
ing a head on the sixth toss is less than .50. Is he right? Is coin tossing mathematically equiva-
lent to the procedure mentioned in part a or the procedure mentioned in part b? Explain.
4.142 A gambler has four cards — two diamonds and two clubs. The gambler proposes the following game
to you: You will leave the room and the gambler will put the cards face down on a table. When you return
to the room, you will pick two cards at random. You will win $10 if both cards are diamonds, you will win
$10 if both are clubs, and for any other outcome you will lose $10. Assuming that there is no cheating,
should you accept this proposition? Support your answer by calculating your probability of winning $10.
184 Chapter 4 Probability
4.143 A thief has stolen Roger's automatic teller machine (ATM) card. The card has a four-digit personal
identification number (PIN). The thief knows that the first two digits are 3 and 5, but he does not know
the last two digits. Thus, the PIN could be any number from 3500 to 3599. To protect the customer, the
automatic teller machine will not allow more than three unsuccessful attempts to enter the PIN. After the
third wrong PIN, the machine keeps the card and allows no further attempts.
a. What is the probability that the thief will find the correct PIN within three tries? (Assume that
the thief will not try the same wrong PIN twice.)
b. If the thief knew that the first two digits were 3 and 5 and that the third digit was either 1 or 7,
what is the probability of the thief guessing the correct PIN in three attempts?
4.144 Consider the following games with two dice.
a. A gambler is going to roll a die four times. If he rolls at least one 6, you must pay him $5. If
he fails to roll a 6 in four tries, he will pay you $5. Find the probability that you must pay the
gambler. Assume that there is no cheating.
b. The same gambler offers to let you roll a pair of dice 24 times. If you roll at least one dou-
ble 6, he will pay you $10. If you fail to roll a double 6 in 24 tries, you will pay him $10.
The gambler says that you have a better chance of winning because your probability of suc-
cess on each of the 24 rolls is 1/36 and you have 24 chances. Thus, he says, your probabil-
ity of winning $10 is 24(1/36) = 2/3. Do you agree with this analysis? If so, indicate why.
If not, point out the fallacy in his argument, and then find the correct probability that you
will win.
4.145 A gambler has given you two jars and 20 marbles. Of these 20 marbles, 10 are red and 10 are green.
You must put all 20 marbles in these two jars in such a way that each jar must have at least one marble
in it. Then a friend of yours, who is blindfolded, will select one of the two jars at random and then will
randomly select a marble from this jar. If the selected marble is red, you and your friend win $100.
a. If you put 5 red marbles and 5 green marbles in each jar, what is the probability that your friend
selects a red marble?
b. If you put 2 red marbles and 2 green marbles in one jar and the remaining marbles in the other
jar, what is the probability that your friend selects a red marble?
c How should these 20 marbles be distributed among the two jars in order to give your friend the
highest possible probability of selecting a red marble?
4.146 A screening test for a certain disease is prone to giving false positives or false negatives. If a
patient being tested has the disease, the probability that the test indicates a (false) negative is .13. If the
patient does not have the disease, the probability that the test indicates a (false) positive is .10. Assume
that 3% of the patients being tested actually have the disease. Suppose that one patient is chosen at
random and tested. Find the probability that
a. this patient has the disease and tests positive
b. this patient does not have the disease and tests positive
c. this patient tests positive
d. this patient has the disease given that he or she tests positive
(Hint: A tree diagram may be helpful in part c.)
4.147 A pizza parlor has 12 different toppings available for its pizzas, and 2 of these toppings are pep-
peroni and anchovies. If a customer picks 2 toppings at random, find the probability that
a. neither topping is anchovies
b. pepperoni is one of the toppings
4.148 An insurance company has information that 93% of its auto policy holders carry collision cover-
age or uninsured motorist coverage on their policies. Eighty percent of the policy holders carry collision
coverage, and 60% have uninsured motorist coverage.
a. What percentage of these policy holders carry both collision and uninsured motorist
coverage?
b. What percentage of these policy holders carry neither collision nor uninsured motorist coverage?
c. What percentage of these policy holders carry collision but not uninsured motorist coverage?
4.149 Many states have a lottery game, usually called a Pick-4, in which you pick a four-digit number
such as 7359. During the lottery drawing, there are four bins, each containing balls numbered through
9. One ball is drawn from each bin to form the four-digit winning number.
a. You purchase one ticket with one four-digit number. What is the probability that you will win
this lottery game?
b. There are many variations of this game. The primary variation allows you to win if the four dig-
its in your number are selected in any order as long as they are the same four digits as obtained
Self-Review Test 185
by the lottery agency. For example, if you pick four digits making the number 1265, then you
will win if 1265, 2615, 5216, 6521, and so forth, are drawn. The variations of the lottery game
depend on how many unique digits are in your number. Consider the following four different
versions of this game.
i. All four digits are unique (e.g., 1234)
ii. Exactly one of the digits appears twice (e.g., 1223 or 9095)
iii. Two digits each appear twice (e.g., 2121 or 5588)
iv. One digit appears three times (e.g., 3335 or 2722)
Find the probability that you will win this lottery in each of these four situations.
4.150 A restaurant chain is planning to purchase 100 ovens from a manufacturer, provided that these ovens
pass a detailed inspection. Because of high inspection costs, 5 ovens are selected at random for inspec-
tion. These 100 ovens will be purchased if at most 1 of the 5 selected ovens fails inspection. Suppose that
there are 8 defective ovens in this batch of 100 ovens. Find the probability that the batch of ovens is pur-
chased. (Note: In Chapter 5 you will learn another method to solve this problem.)
4.151 A production system has two production lines; each production line performs a two-part process,
and each process is completed by a different machine. Thus, there are four machines, which we can iden-
tify as two first-level machines and two second-level machines. Each of the first-level machines works
properly 98% of the time, and each of the second-level machines works properly 96% of the time. All
four machines are independent in regard to working properly or breaking down. Two products enter this
production system, one in each production line.
a. Find the probability that both products successfully complete the two-part process (i.e., all four
machines are working properly).
b. Find the probability that neither product successfully completes the two-part process (i.e., at least
one of the machines in each production line is not working properly).
Self-Review Test
1. The collection of all outcomes for an experiment is called
a. a sample space b. the intersection of events c. joint probability
2. A final outcome of an experiment is called
a. a compound event b. a simple event c. a complementary event
3. A compound event includes
a. all final outcomes
b. exactly two outcomes
c. more than one outcome for an experiment
4. Two equally likely events
a. have the same probability of occurrence
b. cannot occur together
c. have no effect on the occurrence of each other
5. Which of the following probability approaches can be applied only to experiments with equally likely
outcomes?
a. Classical probability b. Empirical probability c. Subjective probability
6. Two mutually exclusive events
a. have the same probability
b. cannot occur together
c. have no effect on the occurrence of each other
7. Two independent events
a. have the same probability b. cannot occur together
c. have no effect on the occurrence of each other
8. The probability of an event is always
a. less than b. in the range to 1.0 c. greater than 1.0
9. The sum of the probabilities of all final outcomes of an experiment is always
a. 100 b. 1.0 c.
186 Chapter 4 Probability
10. The joint probability of two mutually exclusive events is always
a. 1.0 b. between and 1 c.
11. Two independent events are
a. always mutually exclusive
b. never mutually exclusive
c. always complementary
12. A couple is planning their wedding reception. The bride's parents have given them a choice of
four reception facilities, three caterers, five DJs, and two limo services. If the couple randomly selects
one reception facility, one caterer, one DJ, and one limo service, how many different outcomes are
possible?
13. Lucia graduated this year with an accounting degree from Eastern Connecticut State University. She
has received job offers from an accounting firm, an insurance company, and an airline. She cannot de-
cide which of the three job offers she should accept. Suppose she decides to randomly select one of these
three job offers. Find the probability that the job offer selected is
a. from the insurance company
b. not from the accounting firm
14. There are 200 students in a particular graduate program at a state university. Of them, 110 are female
and 125 are out-of-state students. Of the 110 females, 70 are out-of-state students.
a. Are the events "female" and "out-of-state student" independent? Are they mutually exclusive? Ex-
plain why or why not.
b. If one of these 200 students is selected at random, what is the probability that the student
selected is
i. a male?
ii. an out-of-state student given that this student is a female?
15. Reconsider Problem 14. If one of these 200 students is selected at random, what is the probability
that the selected student is a female or an out-of-state student?
16. Reconsider Problem 14. If two of these 200 students are selected at random, what is the probability
that both of them are out-of-state students?
17. The probability that an adult has ever experienced a migraine headache is .35. If two adults are
randomly selected, what is the probability that neither of them has ever experienced a migraine
headache?
18. A hat contains five green, eight red, and seven blue marbles. Let A be the event that a red marble is
drawn if we randomly select one marble out of this hat. What is the probability of A? What is the com-
plementary event of A, and what is its probability?
19. The probability that a randomly selected student from a college is a female is .55 and the probability
that a student works for more than 10 hours per week is .62. If these two events are independent, find the
probability that a randomly selected student is a
a. male and works for more than 10 hours per week
b. female or works for more than 10 hours per week
20. A sample was selected of 506 workers who currently receive two weeks of paid vacation per year.
These workers were asked if they were willing to accept a small pay cut to get an additional week of paid
vacation a year. The following table shows the responses of these workers.
Yes
No
No Response
Man
77
140
32
Woman
104
119
34
a. If one person is selected at random from these 506 workers, find the following probabilities.
i. P(yes)
ii. P(yes | woman)
iii. P{ woman and no)
iv. P(no response or man)
b. Are the events "woman" and "yes" independent? Are they mutually exclusive? Explain why or
why not.
Mini-Projects 187
Mini-Projects
■ MINI-PROJECT 4-1
Suppose that a small chest contains three drawers. The first drawer contains two $1 bills, the second drawer
contains two $100 bills, and the third drawer contains one $1 bill and one $100 bill. Suppose that first a
drawer is selected at random and then one of the two bills inside that drawer is selected at random. We
can define these events:
A = the first drawer is selected B = the second drawer is selected
C = the third drawer is selected D = a $1 bill is selected
a. Suppose when you randomly select one drawer and then one bill from that drawer, the bill you
obtain is a $1 bill. What is the probability that the second bill in this drawer is a $100 bill? In
other words, find the probability P(C \ D) because for the second bill to be $100, it has to be the
third drawer. Answer this question intuitively without making any calculations.
b. Use the relative frequency concept of probability to estimate P(C \ D) as follows. First select
a drawer by rolling a die once. If either 1 or 2 occurs, the first drawer is selected; if either 3
or 4 occurs, the second drawer is selected; and if either 5 or 6 occurs, the third drawer is se-
lected. Whenever C occurs, then select a bill by tossing a coin once. (Note that if either A or
B occurs, then you do not need to toss the coin because each of these drawers contains both
bills of the same denomination.) If you obtain a head, assume that you select a $1 bill; if you
obtain a tail, assume that you select a $100 bill. Repeat this process 100 times. How many
times in these 100 repetitions did the event D occur? What proportion of the time did C occur
when D occurred? Use this proportion to estimate P(C \ D). Does this estimate support your
guess of P(C | D) in part a?
c. Calculate P(C \ D) using the procedures developed in this chapter (a tree diagram may be help-
ful). Was your estimate in part b close to this value? Explain.
■ MINI-PROJECT 4-2
There are two families playing in a park, and each of these two families has two children. The Smith fam-
ily has two daughters, and the Jones family has a daughter and a son. One family is selected at random,
and one of the children from this family is chosen at random.
a. Suppose that the selected child is a girl. What is the probability that the second child is also a girl?
(Note: you need to determine this using conditional probability).
b. Use the relative frequency concept to estimate the probability that the second child in the fam-
ily is a girl given that the selected child is a girl. Use the following process to do so. First toss
a coin to determine whether the Smith family or the Jones family is chosen. If the Smith fam-
ily is selected, then record that the second child in this family is a girl given that the selected
child is a girl. This is so because both children in this family are girls. If the Jones family is
selected, then toss the coin again to select a child and record the gender of the child selected
and that of the second child in this family. Repeat this process 50 times, and then use the re-
sults to estimate the required probability. How close is your estimate to the probability calcu-
lated in part a?
■ MINI-PROJECT 4-3
The dice game Yahtzee® involves five standard dice. On your turn, you can roll all five or fewer dice up
to three times to obtain different sets of numbers on the dice. For example, you will roll all five dice the
first time; if you like two of the five numbers obtained, you can roll the other three dice a second time;
now if you like three of the five numbers obtained, you can roll the other two dice the third time. Some
of the sets of numbers obtained are similar to poker hands (three of a kind, four of a kind, full house, and
so forth). However, a few other hands, like five of a kind (called a yahtzee), are not poker hands (or at
least they are not the hands you would dare to show anyone.)
For the purpose of this project, we will examine the outcomes on the first roll of the five dice. The
five scenarios that we will consider are:
i. Three of a kind — three numbers are the same and the remaining two numbers are both different,
for example, 22254
188 Chapter 4 Probability
ii. Four of a kind — four numbers are the same and the fifth number is different, for example, 44442
iii. Full house — three numbers are the same and the other two numbers are the same, for example,
33366
iv. Large straight — five numbers in a row, for example, 12345
V. Yahtzee — all five numbers are the same, for example, 33333
In the first two cases, the dice that are not part of the three or four of a kind must have different val-
ues than those in the three or four of a kind. For example, 22252 cannot be considered three of a kind,
but 22254 is three of a kind. (Yahtzee players know that this situation differs from the rules of the actual
game, but for the purpose of this project, we will change the rules.)
a. Find the probability for each of these five cases for one roll of the five dice.
b. In a regular game, you do not have to roll all five dice on each of the three rolls. You can leave
some dice on the table and roll the others in an attempt to improve your score. For example, if
you roll 13555 on the first roll of five dice, you can keep the three fives and roll the dice with 1
and 3 outcomes the second time in an attempt to get more fives, or possibly a pair of another num-
ber in order to get a full house. After your second roll, you are allowed to pick up any of the dice
for your third roll. For example, suppose your first roll is 13644 and you keep the two fours. Then
you roll the dice with 1, 3, and 6 outcomes the second time and obtain three fives. Thus, now you
have 55544. Although you met your full house requirement before rolling the dice three times,
you still need a yahtzee. So, you keep the fives and roll the two dice with fours the third time.
Write a paragraph outlining all of the scenarios you will have to consider to calculate the proba-
bility of obtaining a yahtzee within your three rolls.
DECIDE FOR YOURSELF
Deciding About Production Processes
Henry Ford was one of the major developers of mass production.
Imagine if his factory had only one production line! If any compo-
nent in that production line would have broken down, all production
would have come to a halt. In order for mass production to be suc-
cessful, the factory must be able to continue production when one or
more machines in the production process breakdown. Automobile
factories, like many other forms of production, have multiple pro-
duction lines running side by side. So if one production line is shut
down due to a breakdown, the other production lines can still oper-
ate. Probability theory can be used to study the reliability of produc-
tion systems by determining the likelihood that a system will contin-
ue to operate even when some parts of the system fail.
In order to study such systems, we have to consider how they
are set up. These systems comprise two types of arrangements: series
and parallel. In a series system, a process is sequential. One part of
the process must be completed before the item can move to the next
part of the process. If any part of the system breaks down, none of
the tasks that follow can be completed. In the auto example, if some-
thing in a series system breaks down while the chassis is being con-
structed, it will be impossible to install the seats, the windshield, the
engine, and so forth.
In a parallel system, various processes work side by side. In
some cases the processes are like toll collectors at a bridge or on a
highway. As long as there is at least one toll collector working,
traffic will continue moving, although more toll collectors would
certainly speed up the process. In a computer network, different
servers are set up in parallel systems. If one server (such as the
e-mail server) goes down, people on the network can still access
the Web and file servers. However, if the servers were set up in a
series system and the e-mail server failed, nobody would be able to
do anything.
Let us consider a simplified example. Suppose a production
line involves five tasks. Each of the machines that perform these
tasks works successfully 97% of the time. In other words, the
probability that a specific task can be completed (without interrup-
tion) is .97. For the sake of simplicity, let us assume that the
machines work and fail independently of each other. Furthermore,
suppose that the factory has three of these lines running in a par-
allel system. Following are some of the questions that arise. Act as
if you are in charge of such a production process and try to answer
these questions.
1. What is the probability that all five tasks in a single line are com-
pleted without interruption?
2. What is the probability that at least one of the three production
lines is working properly?
3. Why is the probability that a specific line works properly lower
than the probability that at least one of the lines in the factory works
properly?
4. What happens to the reliability of the system if an additional task
is added to each line?
5. What happens when the number of tasks remains constant, but
another line is added?
Technology Instruction 189
ECHNOLOGY
INSTRUCTION
Generating Random Numbers
1,50)-M_1
Screen 4.1
1. To generate a random number (not necessarily an integer) uniformly distributed between m
and n, select MATH >PRB and type rancT(n— m)+m.
2. To generate a random number that is an integer uniformly distributed between m and n, se-
lect MATH >PRB and type randlnt(m,n).
3. To create a sequence of random numbers (integer or noninteger) and store them in a list,
you will need to use the seq( function in conjunction with the appropriate random number
function from step 1 or step 2. Specifically, select 2nd >STAT >OPS >SEQ(, then type
the function from step 1 or step 2, then type ,X,1> quantity of random numbers you
want) >STO— > >L1 >ENTER. These instructions will store the data in list LI (see
Screen 4.1). However, you can replace LI by any other list you want in the above
instructions.
Uniform Distribution
Generate 1 1
Store in column(s):
rows of data
cl
Lower endpoint: fi
Upper endpoint: |ioo
Select
Help
OK
Cancel
1. To generate random numbers (not necessarily
integers) uniformly distributed between m and
n, select Calc >Random Data >Uniform.
Enter the number of rows of data, the column
where you wish to store the data, and the
minimum m and maximum n values for the
numbers (see Screens 4.2
and 4.3).
To generate random inte-
gers uniformly distributed
between m and n, select
Calc >Random Data
> Integer. Enter the num-
ber of rows of data, the
column where you wish to
store the data, and the
minimum m and maximum
n values for the integers.
4-
C1
1
82.8024
2
6.5293
3
77.9912
4
2.4420
5
61.5157
6
2.0113
7
52.3235
8
63.2997
9
83.7044
10
26.6449
Screen 4.2
Screen 4.3
1. To generate a random number (not necessarily an integer) uniformly distributed between
m and n, enter the formula =rand()*(n— m)+m. If you need more than one random
number, copy and paste the formula into as many cells as you need. The numbers will
be recalculated every time any cell in the spreadsheet is calculated or recalculated (see
Screen 4.4).
A
B
1
=RAND(f(100-1)+1
2
Screen 4.4
190 Chapter 4 Probability
2. To generate a random integer uniformly distributed between m and re, enter the formula
=floor(rand()*(n— m+l)+m,l). If you need more than one random number, copy and
paste the formula into as many cells as you need. The numbers will be recalculated every
time any cell in the spreadsheet is calculated or recalculated.
3. To enter a random number (either type) that stays fixed after it is calculated, select the cell
containing the formula, select the formula bar, and press F9. (Note that this procedure
works only one cell at a time.) If you have a bunch of random numbers that you wish to
keep fixed after being calculated, highlight all of the numbers, select Edit >Copy, go to an
empty column, then select Edit >Paste Special and check the Values box.
TECHNOLOGY ASSIGNMENTS
TA4.1 You want to simulate the tossing of a coin. Assign a value of (zero) to Head and a value of 1
to Tail.
a. Simulate 50 tosses of the coin by generating 50 random (integer) numbers between and 1. Then
calculate the mean of these 50 numbers. This mean gives you the proportion of 50 tosses that resulted
in tails. Using this proportion, calculate the number of heads and tails you obtained in 50 simulated
tosses.
b. Repeat part a by simulating 600 tosses.
c. Repeat part a by simulating 4000 tosses.
Comment on the percentage of Tails obtained as the number of tosses is increased.
TA4.2 You want to simulate the rolling of a die. Assign the values 1 through 6 to the outcomes from
1-spot through 6-spots on the die, respectively.
a. Simulate 200 rolls of the die by generating 200 random (integer) numbers between 1 and 6. Then
make a histogram for these 200 numbers.
b. Repeat part a by simulating 1000 rolls of the die.
c. Repeat part a by simulating 6000 rolls of the die.
Comment on the histograms obtained in parts a through c.
TA4.3 Random number generators can be used to simulate the behavior of many different types of events,
including those that have an infinite set of possibilities.
a. Generate a set of 200 random numbers on the interval to 1 and save them to a column or list in
the technology you are using.
b. Generate a second set of 200 random numbers, but on the interval 12.3 to 13.3 and save them to
a different column or list in the technology you are using.
c. Create histograms of the simulated data in each of the two columns for parts a and b. Compare
the shapes of the histograms.
Discrete Random Variables and
Their Probability Distributions
Now that you know a little about probability, do you feel lucky? If you've got $20 to spend on
lunch today, are you willing to spend it all on twenty SI lotto tickets to increase your chances of
winning? What if you know that, depending on what state you are in, it could mean buying as
many as 18 million SI tickets to cover all the possible combinations to have a definite chance to
win? (See Case Study 5-2.) That is a lot of lunch! How do we go about determining the outcomes
and probabilities in a lotto game?
Chapter 4 discussed the concepts and rules of probability. This chapter extends the concept of prob-
ability to explain probability distributions. As we saw in Chapter 4, any given statistical experiment has
more than one outcome. It is impossible to predict which of the many possible outcomes will occur
if an experiment is performed. Consequently, decisions are made under uncertain conditions. For ex-
ample, a lottery player does not know in advance whether or not he is going to win that lottery. If he
knows that he is not going to win, he will definitely not play. It is the uncertainty about winning (some
positive probability of winning) that makes him play. This chapter shows that if the outcomes and their
probabilities for a statistical experiment are known, we can find out what will happen, on average, if
that experiment is performed many times. For the lottery example, we can find out what a lottery
player can expect to win (or lose), on average, if he continues playing this lottery again and again.
In this chapter, random variables and types of random variables are explained. Then, the concept
of a probability distribution and its mean and standard deviation for a discrete random variable
are discussed. Finally, three special probability distributions for a discrete random variable— the
binomial probability distribution, the hypergeometric probability distribution, and the Poisson probability
distribution— are developed.
5.1 Random Variables
5.2 Probability Distribution
of a Discrete Random
Variable
5.3 Mean of a Discrete
Random Variable
5.4 Standard Deviation of a
Discrete Random
Variable
Case Study 5-1 Aces High
Instant Lottery Game—
20th Edition
5.5 Factorials,
Combinations, and
Permutations
Case Study 5-2
Lotto
Playing
5.6 The Binomial Probability
Distribution
5.7 The Hypergeometric
Probability Distribution
5.8 The Poisson Probability
Distribution
Case Study 5-3 Ask Mr.
Statistics
Case Study 5-4 Living and
Dying in the USA
191
192 Chapter 5 Discrete Random Variables and Their Probability Distributions
5.1 Random Variables
Suppose Table 5.1 gives the frequency and relative frequency distributions of the number of ve-
hicles owned by all 2000 families living in a small town.
Table 5.1 Frequency and Relative Frequency Distributions of the
Number of Vehicles Owned by Families
Number of Vehicles Owned
Frequency
Relative Frequency
30
30/2000 = .015
1
470
470/2000 = .235
2
850
850/2000 = .425
3
490
490/2000 = .245
4
160
160/2000 = .080
jV = 2000
Sum = 1.000
Suppose one family is randomly selected from this population. The process of randomly
selecting a family is called a random or chance experiment. Let x denote the number of vehi-
cles owned by the selected family. Then x can assume any of the five possible values (0, 1, 2,
3, and 4) listed in the first column of Table 5.1. The value assumed by x depends on which fam-
ily is selected. Thus, this value depends on the outcome of a random experiment. Consequently,
x is called a random variable or a chance variable. In general, a random variable is denoted
by x or y.
Definition
Random Variable A random variable is a variable whose value is determined by the outcome
of a random experiment.
As will be explained next, a random variable can be discrete or continuous.
5.1.1 Discrete Random Variable
A discrete random variable assumes values that can be counted. In other words, the consec-
utive values of a discrete random variable are separated by a certain gap.
Definition
Discrete Random Variable A random variable that assumes countable values is called a discrete
random variable.
In Table 5.1, the number of vehicles owned by a family is an example of a discrete random
variable because the values of the random variable x are countable: 0, 1,2, 3, and 4.
Here are some other examples of discrete random variables:
1. The number of cars sold at a dealership during a given month
2. The number of houses in a certain block
3. The number of fish caught on a fishing trip
4. The number of complaints received at the office of an airline on a given day
5.1 Random Variables 193
5. The number of customers who visit a bank during any given hour
6. The number of heads obtained in three tosses of a coin
5.1.2 Continuous Random Variable
A random variable whose values are not countable is called a continuous random variable. A
continuous random variable can assume any value over an interval or intervals.
Definition
Continuous Random Variable A random variable that can assume any value contained in one or
more intervals is called a continuous random variable.
Because the number of values contained in any interval is infinite, the possible number of
values that a continuous random variable can assume is also infinite. Moreover, we cannot count
these values. Consider the life of a battery. We can measure it as precisely as we want. For in-
stance, the life of this battery may be 40 hours, or 40.25 hours, or 40.247 hours. Assume that
the maximum life of a battery is 200 hours. Let x denote the life of a randomly selected bat-
tery of this kind. Then, x can assume any value in the interval to 200. Consequently, x is a
continuous random variable. As shown in the diagram, every point on the line representing the
interval to 200 gives a possible value of x.
— I
200
Every point on this line represents a possible value of x that denotes the
life of a battery. There is an infinite number of points on this line. The
values represented by points on this line are uncountable.
The following are some examples of continuous random variables:
1. The length of a room
2. The time taken to commute from home to work
3. The amount of milk in a gallon (note that we do not expect "a gallon" to contain exactly
one gallon of milk but either slightly more or slightly less than one gallon).
4. The weight of a fish
5. The price of a house
This chapter is limited to a discussion of discrete random variables and their probability
distributions. Continuous random variables will be discussed in Chapter 6.
EXERCISES
CONCEPTS AND PROCEDURES
5.1 Explain the meaning of a random variable, a discrete random variable, and a continuous random vari-
able. Give one example each of a discrete random variable and a continuous random variable.
5.2 Classify each of the following random variables as discrete or continuous.
a. The time left on a parking meter
b. The number of bats broken by a major league baseball team in a season
c. The number of cars in a parking lot
d. The total pounds of fish caught on a fishing trip
e. The number of cars crossing a bridge on a given day
f. The time spent by a physician examining a patient
194 Chapter 5 Discrete Random Variables and Their Probability Distributions
5.3 Indicate which of the following random variables are discrete and which are continuous.
a. The number of new accounts opened at a bank during a certain month
b. The time taken to run a marathon
c. The price of a concert ticket
d. The number of times a person says "please" in a day
e. The points scored in a football game
f. The weight of a randomly selected package
■ APPLICATIONS
5.4 A household can watch news on any of the three networks — ABC, CBS, or NBC. On a certain day,
five households randomly and independently decide which channel to watch. Let x be the number of house-
holds among these five that decide to watch news on ABC. Is x a discrete or a continuous random vari-
able? Explain.
5.5 One of the four gas stations located at an intersection of two major roads is a Texaco station. Sup-
pose the next six cars that stop at any of these four gas stations make their selections randomly and
independently. Let x be the number of cars in these six that stop at the Texaco station. Is x a discrete or a
continuous random variable? Explain.
5.2 Probability Distribution of a Discrete
Random Variable
Let x be a discrete random variable. The probability distribution of x describes how the prob-
abilities are distributed over all the possible values of x.
Definition
Probability Distribution of a Discrete Random Variable The probability distribution of a discrete
random variable lists all the possible values that the random variable can assume and their cor-
responding probabilities.
Example 5-1 illustrates the concept of the probability distribution of a discrete random
variable.
■ EXAMPLE 5-1
Recall the frequency and relative frequency distributions of the number of vehicles owned by
families given in Table 5.1. That table is reproduced as Table 5.2. Let x be the number of
vehicles owned by a randomly selected family. Write the probability distribution of x.
Table 5.2 Frequency and Relative Frequency
Distributions of the Number of Vehicles
Owned by Families
Number of
Vehicles Owned
Frequency
Relative
Frequency
30
.015
1
470
.235
2
850
.425
3
490
.245
4
160
.080
N = 2000 Sum = 1.000
Writing the probability
distribution of a discrete
random variable.
5.2 Probability Distribution of a Discrete Random Variable 195
Solution In Chapter 4, we learned that the relative frequencies obtained from an experiment
or a sample can be used as approximate probabilities. However, when the relative frequencies
represent the population, as in Table 5.2, they give the actual (theoretical) probabilities of out-
comes. Using the relative frequencies of Table 5.2, we can write the probability distribution
of the discrete random variable x in Table 5.3.
Table 5.3 Probability Distribution of the Number
of Vehicles Owned by Families
Number of Vehicles Owned
X
Probability
P(x)
.015
1
.235
2
.425
3
.245
4
.080
%P(x) = 1.000
The probability distribution of a discrete random variable possesses the following two
characteristics:
1. The probability assigned to each value of a random variable x lies in the range to 1; that
is, ^ P(x) < 1 for each x.
2. The sum of the probabilities assigned to all possible values of x is equal to 1.0; that is,
= 1. (Remember, if the probabilities are rounded, the sum may not be exactly 1.0.)
Two Characteristics of a Probability Distribution The probability distribution of a discrete ran-
dom variable possesses the following two characteristics.
1. ^ P(x) < 1 for each value of x
2. XPix) = 1
These two characteristics are also called the two conditions that a probability distribution
must satisfy. Notice that in Table 5.3 each probability listed in the column labeled P(x) is be-
tween and 1. Also, ~%P(x) = 1.0. Because both conditions are satisfied, Table 5.3 represents
the probability distribution of x.
From Table 5.3, we can read the probability for any value of x. For example, the probabil-
ity that a randomly selected family from this town owns two vehicles is .425. This probability
is written as
P (x = 2) = .425 or P(2) = .425
The probability that the selected family owns more than two vehicles is given by the sum
of the probabilities of owning three and four vehicles. This probability is .245 + .080 = .325,
which can be written as
P(x > 2) = P(x = 3) + P(x = 4) = P(3) + P(4) = .245 + .080 = .325
The probability distribution of a discrete random variable can be presented in the form of
a mathematical formula, a table, or a graph. Table 5.3 presented the probability distribution in
tabular form. Figure 5.1 shows the graphical presentation of the probability distribution of Table
5.3. In this figure, each value of x is marked on the horizontal axis. The probability for each
196 Chapter 5 Discrete Random Variables and Their Probability Distributions
Figure 5.T Graphical presentation of the probability distribution of P(x)
Table 5.3. 50Q
.400
.300
.200
.100
n .
1
value of x is exhibited by the height of the corresponding bar. Such a graph is called a bar
graph. This section does not discuss the presentation of a probability distribution using a math-
ematical formula.
EXAMPLE 5-2
Verifying the conditions of a
probability distribution.
Each of the following tables lists certain values of x and their probabilities. Determine whether
or not each table represents a valid probability distribution.
(a) x
P(x)
(b) x
P(x)
(c) X
P(x)
.08
2
.25
7
.70
1
.11
3
.34
8
.50
2
.39
4
.28
9
-.20
3
.27
5
.13
Solution
(a) Because each probability listed in this table is in the range to 1, it satisfies the
first condition of a probability distribution. However, the sum of all probabilities
is not equal to 1.0 because %P(x) = .08 + .11 + .39 + .27 = .85. Therefore, the
second condition is not satisfied. Consequently, this table does not represent a valid
probability distribution.
(b) Each probability listed in this table is in the range to 1. Also, %P(x) = .25 + .34 +
.28 + .13 = 1.0. Consequently, this table represents a valid probability distribution.
(c) Although the sum of all probabilities listed in this table is equal to 1.0, one of the
probabilities is negative. This violates the first condition of a probability distribution.
Therefore, this table does not represent a valid probability distribution. ■
■ EXAMPLE 5-3
The following table lists the probability distribution of the number of breakdowns per week
for a machine based on past data.
Breakdowns per week
1
2
3
Probability
.15
.20
.35
.30
(a) Present this probability distribution graphically.
(b) Find the probability that the number of breakdowns for this machine during a given
week is
i. exactly 2 ii. to 2
iii. more than 1 iv. at most 1
5.2 Probability Distribution of a Discrete Random Variable
Solution Let x denote the number of breakdowns for this machine during a given week.
Table 5.4 lists the probability distribution of x.
Table 5.4
Probability
Is lr) LI 1 UULlUll \JL
the Number of
Breakdowns
X
P(x)
.15
1
.20
2
.35
3
.30
%P(x) = 1.00
(a) Figure 5.2 shows the bar graph of the probability distribution of Table 5.4.
P(x),
.40
Graphing a probability
distribution.
Figure 5.2 Graphical presentation of
the probability distribution of Table 5.4.
.30
.20
.10
1 2 3 x
(b) Using Table 5.4, we can calculate the required probabilities as follows.
Finding the probabilities
i. The probability of exactly two breakdowns is of events for a discrete
^(exactly 2 breakdowns) = P(x = 2) = .35 random variable.
ii. The probability of to 2 breakdowns is given by the sum of the probabilities of
0, 1, and 2 breakdowns:
P(0 to 2 breakdowns) = P(0 < x < 2)
= P(x = 0) + P(x = 1) + P(x = 2)
= .15 + .20 + .35 = .70
iii. The probability of more than 1 breakdown is obtained by adding the probabilities
of 2 and 3 breakdowns:
^(more than 1 breakdown) = P(x > 1)
= P(x = 2) + P(x = 3)
= .35 + .30 = .65
iv. The probability of at most 1 breakdown is given by the sum of the probabilities
of and 1 breakdown:
P(at most 1 breakdown) = P(x ^1)
= P(x = 0) + P(x = 1)
= .15 + .20 = .35 ■
198 Chapter 5 Discrete Random Variables and Their Probability Distributions
EXAMPLE 5-4
According to a survey, 60% of all students at a large university suffer from math anxiety. Two
students are randomly selected from this university. Let x denote the number of students in
probability distribution. , . , , . , . _ , , .
this sample who suffer from math anxiety. Develop the probability distribution of x.
Solution Let us define the following two events:
N = the student selected does not suffer from math anxiety
M = the student selected suffers from math anxiety
As we can observe from the tree diagram of Figure 5.3, there are four possible outcomes
for this experiment: AW (neither of the students suffers from math anxiety), NM (the first
student does not suffer from math anxiety and the second does), MN (the first student suf-
fers from math anxiety and the second does not), and MM (both students suffer from math
anxiety). The probabilities of these four outcomes are listed in the tree diagram. Because
60% of the students suffer from math anxiety and 40% do not, the probability is .60 that
any student selected suffers from math anxiety and .40 that he or she does not.
First student
Second student
Final outcomes and
their probabilities
P(NN) = (.40) (.40) = .16
P{NM) = (.40) (.60) = .24
P(MN) = (.60) (.40) = .24
P(MM) = (.60) (.60) = .36
Figure 5.3 Tree diagram.
In a sample of two students, the number who suffer from math anxiety can be (AW),
1 (AM or MN), or 2 (MM). Thus, x can assume any of three possible values: 0, 1, or 2. The
probabilities of these three outcomes are calculated as follows:
P(x = 0) = P(AW) = .16
P(x = 1) = P(NM or MN) = P(NM) + P(MN) = .24 + .24 = .48
P(x = 2) = P(MM) = .36
Using these probabilities, we can write the probability distribution of x as in Table 5.5.
Table 5.5 Probability Distribution of the Number of Students
with Math Anxiety in a Sample of Two Students
X
P(x)
.16
1
.48
2
.36
tP(x) = 1.00
5.2 Probability Distribution of a Discrete Random Variable 199
EXERCISES
CONCEPTS AND PROCEDURES
5.6 Explain the meaning of the probability distribution of a discrete random variable. Give one example
of such a probability distribution. What are the three ways to present the probability distribution of a dis-
crete random variable?
5.7 Briefly explain the two characteristics (conditions) of the probability distribution of a discrete ran-
dom variable.
5.8 Each of the following tables lists certain values of x and their probabilities. Verify whether or not
each represents a valid probability distribution.
X
P(x)
b. x
P(x)
c. X
P(x)
.10
2
.35
7
-.25
1
.05
3
.28
8
.85
2
.45
4
.20
9
.40
3
.40
5
.14
5.9 Each of the following tables lists certain values of x and their probabilities. Determine whether or
not each one satisfies the two conditions required for a valid probability distribution.
X
P(x)
b. x
P(x)
c. X
P(x)
5
-.36
1
.27
.15
6
.48
2
.24
1
.08
7
.62
3
.49
2
.20
8
.26
3
.50
5.10 The following table gives the probability distribution of a discrete random variable x.
X
1
2
3
4
5
6
P(x)
.11
.19
.28
.15
.12
.09
.06
Find the following probabilities.
a. P(x = 3) b. P(x < 2) c. P(x > 4) d. P(l < x < 4)
e. Probability that x assumes a value less than 4
f. Probability that x assumes a value greater than 2
g. Probability that x assumes a value in the interval 2 to 5
5.11 The following table gives the probability distribution of a discrete random variable x.
X
1
2
3
4
5
PQc)
.03
.17
.22
.31
.15
.12
Find the following probabilities.
a. P(x = 1) b. P(x < 1) c. P(x > 3) d. P(0 < x < 2)
e. Probability that x assumes a value less than 3
f. Probability that x assumes a value greater than 3
g. Probability that x assumes a value in the interval 2 to 4
■ APPLICATIONS
5.12 A review of emergency room records at rural Millard Fellmore Memorial Hospital was performed
to determine the probability distribution of the number of patients entering the emergency room during a
1-hour period. The following table lists the distribution.
200 Chapter 5 Discrete Random Variables and Their Probability Distributions
Patients per hour
1
2
3
4
5
6
Probability
.2725
.3543
.2303
.0998
.0324
.0084
.0023
a. Graph the probability distribution.
b. Determine the probability that the number of patients entering the emergency room during a ran-
domly selected 1-hour period is
i. 2 or more ii. exactly 5 iii. fewer than 3 iv. at most 1
5.13 Nathan Cheboygan, a singing gambler from northern Michigan, is famous for his loaded dice. The
following table shows the probability distribution for the sum, denoted by x, of the faces on a pair of
Nathan's dice.
X
2
3
4
5
6
7
8
9
10
11
12
P(x)
.065
.065
.08
.095
.11
.17
.11
.095
.08
.065
.065
a. Draw a bar graph for this probability distribution.
b. Determine the probability that the sum of the faces on a single roll of Nathan's dice is
i. an even number ii. 7 or 1 1 iii. 4 to 6 iv. no less than 9
5.14 The H2 Hummer limousine has eight tires on it. A fleet of 1300 H2 limos was fit with a batch of
tires that mistakenly passed quality testing. The following table lists the frequency distribution of the num-
ber of defective tires on the 1300 H2 limos.
Number of defective tires
1
2
3
4
5
6
7
8
Number of H2 limos
59
224
369
347
204
76
18
2
1
a. Construct a probability distribution table for the numbers of defective tires on these limos. Draw
a bar graph for this probability distribution.
b. Are the probabilities listed in the table of part a exact or approximate probabilities of the various
outcomes? Explain.
c. Let x denote the number of defective tires on a randomly selected H2 limo. Find the following
probabilities.
i. P(x = 0) ii. P (x < 4) iii. P(3 < x < 7) iv. P(x > 2)
5.15 One of the most profitable items at Al's Auto Security Shop is the remote starting system. Let x be
the number of such systems installed on a given day at this shop. The following table lists the frequency
distribution of x for the past 80 days.
X
1
2
3
4
5
f
8
20
24
16
12
a. Construct a probability distribution table for the number of remote starting systems installed on
a given day. Draw a graph of the probability distribution.
b. Are the probabilities listed in the table of part a exact or approximate probabilities of various out-
comes? Explain.
c. Find the following probabilities.
i. P(x = 3) ii. P(x a 3) iii. P(2 < x < 4) iv. P(x < 4)
5.16 Five percent of all cars manufactured at a large auto company are lemons. Suppose two cars are se-
lected at random from the production line of this company. Let x denote the number of lemons in this sam-
ple. Write the probability distribution of x. Draw a tree diagram for this problem.
5.17 In the 2008 Beach to Beacon 10K run, 27.4% of the 5248 participants finished the race in 49 min-
utes 42 seconds (49:42) or faster, which is equivalent to a pace of less than 8 minutes per mile (Source:
http://www.beach2beacon.org/b2b_2008_runners.htm). Suppose that this result holds true for all people
who would participate in and finish a 10K race. Suppose that two 10K runners are selected at random.
Let x denote the number of runners in these two who would finish a 10K race in 49:42 or less. Construct
the probability distribution table of x. Draw a tree diagram for this problem.
5.18 According to a survey, 30% of adults are against using animals for research. Assume that this result
holds true for the current population of all adults. Let x be the number of adults who are against using an-
imals for research in a random sample of two adults. Obtain the probability distribution of x. Draw a tree
diagram for this problem.
5.3 Mean of a Discreate Random Variable 201
5.19 In a 2006 ABC News poll, 37% of adult Americans stated that they encounter "rude and disrespectful
behavior" often {Source: http://abcnews.go.com/images/Politics/1005alHowRude.pdf). Suppose that this re-
sult holds true for the current population of adult Americans. Suppose that two adult Americans are selected
at random. Let x denote the number of adult Americans in these two who encounter "rude and disrespectful
behavior" often. Construct the probability distribution table of x. Draw a tree diagram for this problem.
*5.20 In a group of 12 persons, 3 are left-handed. Suppose that 2 persons are randomly selected from this
group. Let x denote the number of left-handed persons in this sample. Write the probability distribution
of x. You may draw a tree diagram and use it to write the probability distribution. (Hint: Note that the
selections are made without replacement from a small population. Hence, the probabilities of outcomes
do not remain constant for each selection.)
*5.21 In a group of 20 athletes, 6 have used performance-enhancing drugs that are illegal. Suppose that
2 athletes are randomly selected from this group. Let x denote the number of athletes in this sample who
have used such illegal drugs. Write the probability distribution of x. You may draw a tree diagram and use
that to write the probability distribution. (Hint: Note that the selections are made without replacement from
a small population. Hence, the probabilities of outcomes do not remain constant for each selection.)
5.3 Mean of a Discrete Random Variable
The mean of a discrete random variable, denoted by /a, is actually the mean of its probabil-
ity distribution. The mean of a discrete random variable x is also called its expected value and
is denoted by E(x). The mean (or expected value) of a discrete random variable is the value that
we expect to observe per repetition, on average, if we perform an experiment a large number
of times. For example, we may expect a car salesperson to sell, on average, 2.4 cars per week.
This does not mean that every week this salesperson will sell exactly 2.4 cars. (Obviously one
cannot sell exactly 2.4 cars.) This simply means that if we observe for many weeks, this sales-
person will sell a different number of cars during different weeks; however, the average for all
these weeks will be 2.4 cars per week.
To calculate the mean of a discrete random variable x, we multiply each value of x by the
corresponding probability and sum the resulting products. This sum gives the mean (or expected
value) of the discrete random variable x.
Mean of a Discrete Random Variable The mean of a discrete random variable x is the value that
is expected to occur per repetition, on average, if an experiment is repeated a large number of
times. It is denoted by i± and calculated as
/JL = 2xP(x)
The mean of a discrete random variable x is also called its expected value and is denoted by
E(x); that is,
E(x) = XxP(x)
Example 5-5 illustrates the calculation of the mean of a discrete random variable.
■ EXAMPLE 5-5
Recall Example 5-3 of Section 5.2. The probability distribution Table 5.4 from that exam-
ple is reproduced below. In this table, x represents the number of breakdowns for a machine
during a given week, and P(x) is the probability of the corresponding value of x.
x P(x)
.15
1
.20
2
.35
3
.30
Find the mean number of breakdowns per week for this machine.
Calculating and interpreting
the mean of a discrete random
variable.
202
Chapter 5 Discrete Random Variables and Their Probability Distributions
Solution To find the mean number of breakdowns per week for this machine, we multiply
each value of x by its probability and add these products. This sum gives the mean of the
probability distribution of x. The products xP(x) are listed in the third column of Table 5.6.
The sum of these products gives XxP(x), which is the mean of x.
Table 5.6 Calculating the Mean for the Probability
Distribution of Breakdowns
X
P(x)
xP(x)
.15
0(.15) =
.00
1
.20
1(.20) =
.20
2
.35
2(.35) =
.70
3
.30
3(.30) =
.90
%xP(x) =1.80
The mean is
IX = ZxP(x) = 1.80
Thus, on average, this machine is expected to break down 1.80 times per week over a pe-
riod of time. In other words, if this machine is used for many weeks, then for certain weeks
we will observe no breakdowns; for some other weeks we will observe one breakdown per
week; and for still other weeks we will observe two or three breakdowns per week. The
mean number of breakdowns is expected to be 1.80 per week for the entire period.
Note that /jl = 1.80 is also the expected value of x. It can also be written as
E(x) = 1.80 ■
Case Study 5-1 illustrates the calculation of the mean amount that an instant lottery player
is expected to win.
5.4 Standard Deviation of a Discrete
Random Variable
The standard deviation of a discrete random variable, denoted by cr, measures the spread of
its probability distribution. A higher value for the standard deviation of a discrete random vari-
able indicates that x can assume values over a larger range about the mean. In contrast, a smaller
value for the standard deviation indicates that most of the values that x can assume are clus-
tered closely about the mean. The basic formula to compute the standard deviation of a discrete
random variable is
Cr = VX[(* " /JL) 2 • P( X )]
However, it is more convenient to use the following shortcut formula to compute the standard
deviation of a discrete random variable.
Standard Deviation of a Discrete Random Variable The standard deviation of a discrete random
variable x measures the spread of its probability distribution and is computed as
cr = \Z^x 2 P{x) - jj?
II any ol YOUR CARDS beats
j the DEALER'S CARD y ou win the
WIN UP TO $1,000! PRIZE shown!
Ticket with covered play symbols
ACES HIGH
INSTANT
LOTTERY
GAME-
20TH
EDITION
ncES 5
HIGH
20th H^HgHQIB]
Edition BwM|9g^^l
m^BBH^K It any ot YOUR CARDS beats
aaTi the DEALER'S CARD y ou win the
WIN UP TO $1,000! PRIZE shown! BltTl
Ticket with uncovered play symbols
Note: The Connecticut Lottery encourages responsible play. Purchasers must be
18 or older.
Lottery tickets reproduced with permission of the Connecticut Lottery Corporation.
Currently (2009) the state of Connecticut has in circulation an instant lottery game called Aces High— 20th
Edition, which is one of the longest-running instant tickets lottery game in the state. The cost of each ticket
for this lottery is Si. A player can instantly win $1000, $100, $40, $25, $10, $4, $2, or a free ticket (which
is equivalent to winning $1). Each ticket has six spots to reveal by scratching the latex coating that covers
the play area, one of which contains a dealer's card, four show player's cards, and one indicates the prize
that can be won by the player. A player will win the prize shown in the prize spot if any of the player's four
cards beats the dealer's card.
Based on the information on this lottery game, the following table lists the number of tickets with dif-
ferent prizes in a total of 18,000,000 tickets printed. As is obvious from this table, out of a total of 18,000,000
tickets, 14,176,944 are nonwinning tickets (the ones with a prize of $0 in this table). Of the remaining 3,823,056
tickets with prizes, 1,800,000 have a prize of a free ticket, 1,296,000 have a prize of $2 each, and so forth.
Prize (dollars)
Number of Tickets
14,176,944
Free ticket
1,800,000
2
1,296,000
4
483,480
10
144,000
25
72,000
40
15,552
100
11,880
1000
144
Total = 18,000,000
Source: The Connecticut Lottery Corporation.
The net gain to a player for each of the instant winning tickets is equal to the amount of the prize mi-
nus Si, which is the cost of the ticket. Thus, the net gain for each of the nonwinning tickets is -Si, which
is the cost of the ticket. Let
x = the net amount a player wins by playing this lottery game
The following table shows the probability distribution of x and all the calculations required to com-
pute the mean of x for this probability distribution. The probability of an outcome (net winnings) is calcu-
lated by dividing the number of tickets with that outcome by the total number of tickets.
x (dollars)
P(x)
xP{x)
-1
14,176,944/18,000,000 =
.787608
-.787608
1,800,000/18,000,000 =
.100000
.000000
1
1,296,000/18,000,000 =
.072000
.072000
3
483,480/18,000,000 =
.026860
.080580
9
144,000/18,000,000 =
.008000
.072000
24
72,000/18,000,000 =
.004000
.096000
39
15,552/18,000,000 =
.000864
.033696
99
11,880/18,000,000 =
.000660
.065340
999
144/18,000,000 =
.000008
.007992
2xP(x) = -.36
Hence, the mean or expected value of x is
yu, = 1xP[x) = -S.36
This mean gives the expected value of the random variable x, that is,
E(x) = XxP(x) = -$.36
Thus, the mean of net winnings for this lottery game is —$.36. In other words, all players taken together will
lose an average of $.36 (or 36 cents) per ticket. This can also be interpreted as follows: Only 1 00 - 36 = 64%
of the total money spent by all players on buying lottery tickets for this lottery will be returned to them in
the form of prizes, and 36% will not be returned. (The money that will not be returned to players will cover
the costs of operating the lottery, the commission paid to agents, and revenue to the state of Connecticut.)
Note that the variance a of a discrete random variable is obtained by squaring its standard
deviation.
Example 5-6 illustrates how to use the shortcut formula to compute the standard deviation
of a discrete random variable.
Calculating the standard
deviation of a discrete
random variable.
■ EXAMPLE 5-6
Baier's Electronics manufactures computer parts that are supplied to many computer compa-
nies. Despite the fact that two quality control inspectors at Baier's Electronics check every part
for defects before it is shipped to another company, a few defective parts do pass through these
inspections undetected. Let x denote the number of defective computer parts in a shipment of
400. The following table gives the probability distribution of x.
X
1
2
3
4
5
P(x)
.02
.20
.30
.30
.10
.08
Compute the standard deviation of x.
Solution Table 5.7 shows all the calculations required for the computation of the standard
deviation of x.
204
5.4 Standard Deviation of a Discrete Random Variable 205
Table 5.7 Computations to Find the Standard Deviation
X
/>/,„ \
xr(x)
-2
X
X r%X)
.02
.00
.00
1
.20
.20
1
.20
2
.30
.60
4
1.20
3
.30
.90
9
2.70
4
.10
.40
16
1.60
5
.08
.40
25
2.00
£xP(x) = 2.50
2,x 2 P(x) = 7.70
We perform the following steps to compute the standard deviation of x.
Step 1. Compute the mean of the discrete random variable.
The sum of the products xP(x), recorded in the third column of Table 5.7, gives the
mean of x.
fi = 2xP(x) = 2.50 defective computer parts in 400
Step 2. Compute the value of Xx 2 P(x).
First we square each value of x and record it in the fourth column of Table 5.7. Then
we multiply these values of x 2 by the corresponding values of P(x). The resulting values of
x 2 P(x) are recorded in the fifth column of Table 5.7. The sum of this column is
Xx 2 P(x) = 7.70
Step 3. Substitute the values of fi and 2x 2 P(x) in the formula for the standard deviation
of x and simplify.
By performing this step, we obtain
o- = VSx 2 P(x) - /jl 1 = V7.70 - (2.50) 2 = VT45
= 1.204 defective computer parts
Thus, a given shipment of 400 computer parts is expected to contain an average of 2.50
defective parts with a standard deviation of 1 .204. I
Because the standard deviation of a discrete random variable is obtained by taking the positive -4 Remember
square root, its value is never negative.
■ EXAMPLE 5-7
Loraine Corporation is planning to market a new makeup product. According to the analysis
made by the financial department of the company, it will earn an annual profit of $4.5 million
if this product has high sales, it will earn an annual profit of $1.2 million if the sales are
mediocre, and it will lose $2.3 million a year if the sales are low. The probabilities of these
three scenarios are .32, .51, and .17, respectively.
(a) Let x be the profits (in millions of dollars) earned per annum from this product by the
company. Write the probability distribution of x.
(b) Calculate the mean and standard deviation of x.
206 Chapter 5 Discrete Random Variables and Their Probability Distributions
Solution
Writing the probability
distribution of a discrete
random variable.
(a) The table below lists the probability distribution of x. Note that because x denotes
profits earned by the company, the loss is written as a negative profit in the table.
X
P(x)
4.5
.32
1.2
.51
-2.3
.17
Calculating the mean and
standard deviation of a discrete
random variable.
(b) Table 5.8 shows all the calculations needed for the computation of the mean and stan-
dard deviation of x.
Table 5.8 Computations to Find the Mean and Standard Deviation
X
P(x)
xP(x)
x 2
x 2 P(x)
4.5
.32
1.440
20.25
6.4800
1.2
.51
.612
1.44
.7344
-2.3
.17
-.391
5.29
.8993
2,xP(x) = 1.661
tx 2 P(x) = 8.1137
The mean of x is
[jl = XxP(x) = $1,661 million
The standard deviation of x is
o- = VSjtPO) - i± 2 = V8.1137 - (1.661) 2 = $2,314 million
Thus, it is expected that Loraine Corporation will earn an average of $1,661 million in prof-
its per year from the new product, with a standard deviation of $2,314 million. H
Interpretation of the Standard Deviation
The standard deviation of a discrete random variable can be interpreted or used the same way as
the standard deviation of a data set in Section 3.4 of Chapter 3. In that section, we learned that ac-
cording to Chebyshev's theorem, at least [1 — (l/k 2 )] X 100% of the total area under a curve lies
within k standard deviations of the mean, where k is any number greater than 1 . Thus, if k = 2,
then at least 75% of the area under a curve lies between /jl — 2cr and /jl + 2cr. In Example 5-6,
ix = 2.50 and a = 1.204
Hence,
l± -2a = 2.50 - 2(1.204) = .092
H + 2cr = 2.50 + 2(1.204) = 4.908
Using Chebyshev's theorem, we can state that at least 75% of the shipments (each containing
400 computer parts) are expected to contain .092 to 4.908 defective computer parts each.
EXERCISES
CONCEPTS AND PROCEDURES
5.22 Briefly explain the concept of the mean and standard deviation of a discrete random variable.
5.4 Standard Deviation of a Discrete Random Variable 207
5.23 Find the mean and standard deviation for each of the following probability distributions.
X
P(x)
b. x
P(x)
.16
6
.40
1
.27
7
.26
2
.39
8
.21
3
.18
9
.13
5.24 Find the mean and standard deviation for each of the following probability distributions.
a. x
P(x)
b. x
P(x)
3
.09
.43
4
.21
1
.31
5
.34
2
.17
6
.23
3
.09
7
.13
■ APPLICATIONS
5.25 Let x be the number of errors that appear on a randomly selected page of a book. The following table
lists the probability distribution of x.
x
1
2
3
4
P(x)
.73
.16
.06
.04
.01
Find the mean and standard deviation of x.
5.26 Let x be the number of magazines a person reads every week. Based on a sample survey of adults,
the following probability distribution table was prepared.
x
1
2
3
4
5
P(x)
.36
.24
.18
.10
.07
.05
Find the mean and standard deviation of x.
5.27 The following table gives the probability distribution of the number of camcorders sold on a given
day at an electronics store.
Camcorders sold
1
2
3
4
5
6
Probability
.05
.12
.19
.30
.20
.10
.04
Calculate the mean and standard deviation for this probability distribution. Give a brief interpretation of
the value of the mean.
5.28 The following table, reproduced from Exercise 5.12, lists the probability distribution of the number
of patients entering the emergency room during a 1-hour period at Millard Fellmore Memorial Hospital.
Patients per hour
1
2
3
4
5
6
Probability
.2725
.3543
.2303
.0998
.0324
.0084
.0023
Calculate the mean and standard deviation for this probability distribution.
5.29 Let x be the number of heads obtained in two tosses of a coin. The following table lists the proba-
bility distribution of x.
X
1
2
P(x)
.25
.50
.25
Calculate the mean and standard deviation of x. Give a brief interpretation of the value of the mean.
208
Chapter 5 Discrete Random Variables and Their Probability Distributions
5.30 Let x be the number of potential weapons detected by a metal detector at an airport on a given day.
The following table lists the probability distribution of x.
X
1
2
3
4
5
P(X)
.14
.28
.22
.18
.12
.06
Calculate the mean and standard deviation for this probability distribution and give a brief interpretation
of the value of the mean.
5.31 Refer to Exercise 5.14. Calculate the mean and standard deviation for the probability distribution
you developed for the number of defective tires on all 1300 H2 Hummer limousines. Give a brief inter-
pretation of the values of the mean and standard deviation.
5.32 Refer to Exercise 5.15. Find the mean and standard deviation of the probability distribution you de-
veloped for the number of remote starting systems installed per day by Al's Auto Security Shop over the
past 80 days. Give a brief interpretation of the values of the mean and standard deviation.
5.33 Refer to the probability distribution you developed in Exercise 5.16 for the number of
lemons in two selected cars. Calculate the mean and standard deviation of x for that probability
distribution.
5.34 Refer to the probability distribution developed in Exercise 5.17 for the number of runners in a sam-
ple of two who would finish a 10K race in 49:42 or less. Compute the mean and standard deviation of x
for that probability distribution.
5.35 A contractor has submitted bids on three state jobs: an office building, a theater, and a parking
garage. State rules do not allow a contractor to be offered more than one of these jobs. If this contrac-
tor is awarded any of these jobs, the profits earned from these contracts are $10 million from the of-
fice building, $5 million from the theater, and $2 million from the parking garage. His profit is zero
if he gets no contract. The contractor estimates that the probabilities of getting the office building
contract, the theater contract, the parking garage contract, or nothing are .15, .30, .45, and .10, respec-
tively. Let x be the random variable that represents the contractor's profits in millions of dollars. Write
the probability distribution of x. Find the mean and standard deviation of x. Give a brief interpretation
of the values of the mean and standard deviation.
5.36 An instant lottery ticket costs $2. Out of a total of 10,000 tickets printed for this lottery, 1000
tickets contain a prize of $5 each, 100 tickets have a prize of $10 each, 5 tickets have a prize of
$1000 each, and 1 ticket has a prize of $5000. Let x be the random variable that denotes the net
amount a player wins by playing this lottery. Write the probability distribution of x. Determine the
mean and standard deviation of x. How will you interpret the values of the mean and standard devi-
ation of x?
*5.37 Refer to the probability distribution you developed in Exercise 5.20 for the number of left-
handed persons in a sample of two persons. Calculate the mean and standard deviation of x for that
distribution.
*5.38 Refer to the probability distribution you developed in Exercise 5.21 for the number of athletes in a
random sample of two who have used illegal performance-enhancing drugs. Calculate the mean and stan-
dard deviation of x for that distribution.
5.5 Factorials, Combinations, and Permutations
This section introduces factorials, combinations, and permutations. Of these, factorials and com-
binations will be used in the binomial formula discussed in Section 5.6.
5.5.1 Factorials
The symbol ! (read as factorial) is used to denote factorials. The value of the factorial of a
number is obtained by multiplying all the integers from that number to 1. For example, 7! is
read as "seven factorial" and is evaluated by multiplying all the integers from 7 to 1 .
5.5 Factorials, Combinations, and Permutations 209
Definition
Factorials The symbol n\, read as "n factorial," represents the product of all the integers from
n to 1 . In other words,
n\ = n(n - l)(n - 2)(n - 3) • • ■ 3 • 2 ■ 1
By definition,
0! = 1
■ EXAMPLE 5-8
Evaluate 7!.
Solution To evaluate 7!, we multiply all the integers from 7 to 1
7!=7-6-5-4-3-2-l = 5040
Thus, the value of 7! is 5040.
Evaluating a factorial.
EXAMPLE 5-9
Evaluate 10!. Evaluating a factorial.
Solution The value of 10! is given by the product of all the integers from 10 to 1. Thus,
10! = 10-9-8-7-6-5-4-3-2-l = 3,628,800 ■
■ EXAMPLE 5-10
Evaluate (12-4)!.
Solution The value of (12 - 4)! is
(12 - 4)! = 8! = i
Evaluating a factorial of the
difference between two
numbers.
7-6-5-4-3-2-l = 40,320
■ EXAMPLE 5-11
Evaluate (5-5)!.
Solution The value of (5 - 5)! is 1.
(5 - 5)! = 0! = 1
Note that 0! is always equal to 1.
Statistical software and most calculators can be used to find the values of factorials. Check
if your calculator can evaluate factorials.
5.5.2 Combinations
Quite often we face the problem of selecting a few elements from a large number of distinct
elements. For example, a student may be required to attempt any two questions out of four
in an examination. As another example, the faculty in a department may need to select 3
professors from 20 to form a committee, or a lottery player may have to pick 6 numbers
from 49. The question arises: In how many ways can we make the selections in each of these
examples? For instance, how many possible selections exist for the student who is to
Evaluating a factorial
of zero.
210 Chapter 5 Discrete Random Variables and Their Probability Distributions
choose any two questions out of four? The answer is six. Let the four questions be denoted
by the numbers 1,2, 3, and 4. Then the six selections are
(land 2) (land 3) (land 4) (2 and 3) (2 and 4) (3 and 4)
The student can choose questions 1 and 2, or 1 and 3, or 1 and 4, and so on. Note that in
combinations, all selections are made without replacement.
Each of the possible selections in the above list is called a combination. All six combina-
tions are distinct; that is, each combination contains a different set of questions. It is important
to remember that the order in which the selections are made is not important in the case of com-
binations. Thus, whether we write (1 and 2) or (2 and 1), both these arrangements represent only
one combination.
Definition
Combinations Notation Combinations give the number of ways x elements can be selected from
n elements. The notation used to denote the total number of combinations is
C
which is read as "the number of combinations of n elements selected x at a time."
Suppose there are a total of n elements from which we want to select x elements. Then,
i n denotes the total number of elements
n C x = the number of combinations of n elements selected x at a time
I
1 x denotes the number of elements selected per selection
Number of Combinations The number of combinations for selecting
1 x from n distinct elements
is given by the formula
nl
c =
" A x\(n — x)\
where n\, xl, and (« — x)\ are read as "n factorial," "x factorial,"
and "n minus x factorial,"
respectively.
In the combinations formula,
nl = n(n - l)(n - 2)(n - 3) • ■ ■ 3 ■ 2 • 1
x\ = x(x - 1)(jc - 2) ■ ■ - 3 • 2 • 1
(n — x)l = (n — x)(n — x — 1)(« — x — 2) ■ * ■ 3 • 2 • 1
Note that in combinations, n is always greater than or equal to x. If n is less than x, then
we cannot select x distinct elements from n.
M EXAMPLE 5-12
An ice cream parlor has six flavors of ice cream. Kristen wants to buy two flavors of ice cream.
If she randomly selects two flavors out of six, how many possible combinations are there?
Solution For this example,
n = total number of ice cream flavors = 6
x = number of ice cream flavors to be selected = 2
Finding the number
of combinations using the
formula.
5.5 Factorials, Combinations, and Permutations
Therefore, the number of ways in which Kristen can select two flavors of ice cream out of six is
6! 6! 6 • 5 • 4 • 3 • 2 • 1
15
2!(6 - 2)! 2! 4! 2- 1-4-3 -2-1
Thus, there are 15 ways for Kristen to select two ice cream flavors out of six. ■
■ EXAMPLE 5-13
Three members of a jury will be randomly selected from five people. How many different
combinations are possible?
Solution There are a total of five persons, and we are to select three of them. Hence,
n = 5 and x = 3
Applying the combinations formula, we get
5' 5'
5 C 3
120
3!(5-3)! 3! 2! 6-2
10
If we assume that the five persons are A, B, C, D, and E, then the 10 possible combina-
tions for the selection of three members of the jury are
ABC ABD ABE ACD ACE ADE BCD BCE BDE CDE ■
Finding the number
of combinations and
listing them.
■ EXAMPLE 5-14
Marv & Sons advertised to hire a financial analyst. The company has received applications
from 10 candidates who seem to be equally qualified. The company manager has decided to
call only 3 of these candidates for an interview. If she randomly selects 3 candidates from the
10, how many total selections are possible?
Solution The total number of ways to select 3 applicants from 10 is given by 10 C 3 . Here,
n = 10 and x = 3. We find the number of combinations as follows:
10! 10! 3,628,800
10C3
120
3! (10 -3)! 3! 7! (6) (5040)
Thus, the company manager can select 3 applicants from 10 in 120 ways. H
Statistical software and many calculators can be used to find combinations. Check to see
whether your calculator can do so.
If the total number of elements and the number of elements to be selected are the same,
then there is only one combination. In other words,
C = 1
Also, the number of combinations for selecting zero items from n is 1; that is,
nQ> = 1
For example,
5! 5! 120
Using the combinations
formula.
M Remember
8^0
5!(5 - 5)! 5!0! (120)(1)
8! 8! 40,320
1
0!(8 - 0)! 0! 8! (1)(40,320)
Case Study 5-2 describes the number of ways a lottery player can select six numbers in a
lotto game.
PLAYING
LOTTO
During the past two decades or so, many states have initiated the popular lottery game called lotto. To play
lotto, a player picks any six numbers from a list of numbers usually starting with 1— for example, from 1
through 49. At the end of the lottery period, the state lottery commission randomly selects six numbers
from the same list. If all six numbers picked by a player are the same as the ones randomly selected by
the lottery commission, the player wins.
USA TODAY Snapshots®
A look at statistics that shape the nation
Playing to win &
Source: USA TODAY research By Ron Coddington, USA TODAY
The chart shows the number of combinations (in millions) for picking six numbers for lotto games
played in a few states in 1992. For example, in California a player has to pick six numbers from 1 through
51. As shown in the chart, there are approximately 18 million ways (combinations) to select six numbers
from 1 through 51. In Florida and Massachusetts, a player has to pick six numbers from 1 through 49. For
this lotto, there are approximately 13.9 million combinations.
Let us find the probability that a player who picks six numbers from 1 through 49 wins this game. The
total number of combinations of selecting six numbers from 1 through 49 is obtained as follows:
49!
49 C 6 = = 13,983,816
49 6 6!(49 - 6)!
Thus, there are a total of 13,983,816 different ways to select six numbers from 1 through 49. Hence, the
Source: Chart reprinted with permission probability that a player who plays this lottery once wins is
from USA TODAY, February 27, 1992.
Copyright © 1992, USA TODAY. P(player wins) = 1/13,983,816 = .0000000715
5.5.3 Permutations
The concept of permutations is very similar to that of combinations but with one major difference —
here the order of selection is important. Suppose there are three marbles in a jar — red, green, and
purple — and we select two marbles from these three. When the order of selection is not important,
as we know from the previous section, there are three ways (combinations) to do so. Those three
ways are RG, RP, and GP, where R represents that a red marble is selected, G means a green mar-
ble is selected, and P indicates a purple marble is selected. In these three combinations, the order
of selection is not important, and, thus, RG and GR represent the same selection. However, if the
order of selection is important, then RG and GR are not the same selections, but they are two dif-
ferent selections. Similarly, RP and PR are two different selections, and GP and PG are two
212
5.5 Factorials, Combinations, and Permutations 213
different selections. Thus, if the order in which the marbles are selected is important, then
there are six selections — RG, GR, RR PR, GR and PG. These are called six permutations
or arrangements.
Definition
Permutations Notation Permutations give the total selections of x elements from n (different)
elements in such a way that the order of selections is important. The notation used to denote the
permutations is
P
which is read as "the number of permutations of selecting x elements from n elements."
Permutations are also called arrangements.
Permutations Formula The following formula is used to find the number of permutations or
arrangements of selecting x items out of n items. Note that here, the n items should all be different.
nPx ~ (n - x)\
Example 5-15 shows how to apply this formula.
■ EXAMPLE 5-15
A club has 20 members. They are to select three office holders — president, secretary, and treasurer —
for next year. They always select these office holders by drawing 3 names randomly from the
names of all members. The first person selected becomes the president, the second is the secre-
tary, and the third one takes over as treasurer. Thus, the order in which 3 names are selected from
the 20 names is important. Find the total arrangements of 3 names from these 20.
Solution For this example,
n = total members of the club = 20
x = number of names to be selected = 3
Since the order of selections is important, we find the number of permutations or arrangements
using the following formula:
20!
20!
6840
" 1 (n - x)\ (20 - 3)! 17!
Thus, there are 6840 permutations or arrangements for selecting 3 names out of 20. I
Statistical software and many calculators can find permutations. Check to see whether
your calculator can do it.
Finding the number
of permutations using the
formula.
EXERCISES
CONCEPTS AND PROCEDURES
5.39 Determine the value of each of the following using the appropriate formula.
3! (9 - 3)! 9! (14 - 12)! 5 C 3 V C 4 9 C 3 4 C 3 C 3
5.40 Find the value of each of the following using the appropriate formula.
11!
(7 - 2)! (15 - 5)!
214
Chapter 5 Discrete Random Variables and Their Probability Distributions
■ APPLICATIONS
5.41 A ski patrol unit has nine members available for duty, and two of them are to be sent to rescue an
injured skier. In how many ways can two of these nine members be selected? Now suppose the order of
selection is important. How many arrangements are possible in this case?
5.42 An ice cream shop offers 25 flavors of ice cream. How many ways are there to select 2 different
flavors from these 25 flavors? How many permutations are possible?
5.43 A veterinarian assigned to a racetrack has received a tip that one or more of the 12 horses in the
third race have been doped. She has time to test only 3 horses. How many ways are there to randomly se-
lect 3 horses from these 12 horses? How many permutations are possible?
5.44 An environmental agency will randomly select 4 houses from a block containing 25 houses for a
radon check. How many total selections are possible? How many permutations are possible?
5.45 An investor will randomly select 6 stocks from 20 for an investment. How many total combinations
are possible? If the order in which stocks are selected is important, how many permutations will there be?
5.46 A company employs a total of 16 workers. The management has asked these employees to select
2 workers who will negotiate a new contract with management. The employees have decided to select
the 2 workers randomly. How many total selections are possible? Considering that the order of selection
is important, find the number of permutations.
5.47 In how many ways can a sample (without replacement) of 9 items be selected from a population of
20 items?
5.48 In how many ways can a sample (without replacement) of 5 items be selected from a population of
15 items?
5.6 The Binomial Probability Distribution
The binomial probability distribution is one of the most widely used discrete probability
distributions. It is applied to find the probability that an outcome will occur x times in n per-
formances of an experiment. For example, given that the probability is .05 that a DVD player
manufactured at a firm is defective, we may be interested in finding the probability that in a
random sample of three DVD players manufactured at this firm, exactly one will be defective.
As a second example, we may be interested in finding the probability that a baseball player with
a batting average of .250 will have no hits in 10 trips to the plate.
To apply the binomial probability distribution, the random variable x must be a discrete
dichotomous random variable. In other words, the variable must be a discrete random
variable, and each repetition of the experiment must result in one of two possible outcomes.
The binomial distribution is applied to experiments that satisfy the four conditions of a
binomial experiment. (These conditions are described in Section 5.6.1.) Each repetition of a
binomial experiment is called a trial or a Bernoulli trial (after Jacob Bernoulli). For example,
if an experiment is defined as one toss of a coin and this experiment is repeated 10 times,
then each repetition (toss) is called a trial. Consequently, there are 10 total trials for this
experiment.
5.6.1 The Binomial Experiment
An experiment that satisfies the following four conditions is called a binomial experiment.
1. There are n identical trials. In other words, the given experiment is repeated n times, where
n is a positive integer. All these repetitions are performed under identical conditions.
2. Each trial has two and only two outcomes. These outcomes are usually called a success
and a failure, respectively.
3. The probability of success is denoted by p and that of failure by q, and p + q = 1. The
probabilities p and q remain constant for each trial.
4. The trials are independent. In other words, the outcome of one trial does not affect the outcome
of another trial.
5.6 The Binomial Probability Distribution 215
Conditions of a Binomial Experiment A binomial experiment must satisfy the following four
conditions.
1. There are n identical trials.
2. Each trial has only two possible outcomes.
3. The probabilities of the two outcomes remain constant.
4. The trials are independent.
Note that one of the two outcomes of a trial is called a success and the other a failure.
Notice that a success does not mean that the corresponding outcome is considered favorable or
desirable. Similarly, a failure does not necessarily refer to an unfavorable or undesirable out-
come. Success and failure are simply the names used to denote the two possible outcomes of a
trial. The outcome to which the question refers is usually called a success; the outcome to which
it does not refer is called a failure.
■ EXAMPLE 5-16
Consider the experiment consisting of 10 tosses of a coin. Determine whether or not it is a bi-
nomial experiment.
Solution The experiment consisting of 10 tosses of a coin satisfies all four conditions of a
binomial experiment.
1. There are a total of 10 trials (tosses), and they are all identical. All 10 tosses are per-
formed under identical conditions. Here, n = 10.
2. Each trial (toss) has only two possible outcomes: a head and a tail. Let a head be called
a success and a tail be called a failure.
3. The probability of obtaining a head (a success) is 1/2 and that of a tail (a failure) is
1/2 for any toss. That is,
p = P(H) = 1/2 and q = P(T) = 1/2
The sum of these two probabilities is 1 .0. Also, these probabilities remain the same for
each toss.
4. The trials (tosses) are independent. The result of any preceding toss has no bearing on
the result of any succeeding toss.
Consequently, the experiment consisting of 10 tosses is a binomial experiment. H
■ EXAMPLE 5-17
Five percent of all DVD players manufactured by a large electronics company are defective.
Three DVD players are randomly selected from the production line of this company. The se-
lected DVD players are inspected to determine whether each of them is defective or good. Is
this experiment a binomial experiment?
Solution
1. This example consists of three identical trials. A trial represents the selection of a DVD
player.
2. Each trial has two outcomes: a DVD player is defective or a DVD player is good. Let
a defective DVD player be called a success and a good DVD player be called a failure.
3. Five percent of all DVD players are defective. So, the probability p that a DVD player is
defective is .05. As a result, the probability q that a DVD player is good is .95. These two
probabilities add up to 1 .
Verifying the conditions
of a binomial experiment.
Verifying the conditions
of a binomial experiment.
216 Chapter 5 Discrete Random Variables and Their Probability Distributions
4. Each trial (DVD player) is independent. In other words, if one DVD player is defec-
tive, it does not affect the outcome of another DVD player being defective or good.
This is so because the size of the population is very large compared to the sample size.
Because all four conditions of a binomial experiment are satisfied, this is an example of a
binomial experiment. I
5.6.2 The Binomial Probability Distribution
and Binomial Formula
The random variable x that represents the number of successes in n trials for a binomial
experiment is called a binomial random variable. The probability distribution of x in such ex-
periments is called the binomial probability distribution or simply the binomial distribution.
Thus, the binomial probability distribution is applied to find the probability of x successes in n
trials for a binomial experiment. The number of successes x in such an experiment is a discrete
random variable. Consider Example 5-17. Let x be the number of defective DVD players in a
sample of three. Because we can obtain any number of defective DVD players from zero to
three in a sample of three, x can assume any of the values 0, 1,2, and 3. Since the values of x
are countable, it is a discrete random variable.
Binomial Formula For a binomial experiment, the probability of exactly x successes in n trials
is given by the binomial formula
where
P( X ) = n C xP x q n - x
n = total number of trials
p = probability of success
q = 1 — p = probability of failure
x = number of successes in n trials
n — x = number of failures in n trials
In the binomial formula, n is the total number of trials and x is the total number of successes.
The difference between the total number of trials and the total number of successes, n — x, gives
the total number of failures in n trials. The value of „C A . gives the number of ways to obtain x
successes in n trials. As mentioned earlier, p and q are the probabilities of success and failure,
respectively. Again, although it does not matter which of the two outcomes is called a success
and which a failure, usually the outcome to which the question refers is called a success.
To solve a binomial problem, we determine the values of n, x, n — x, p, and q and then
substitute these values in the binomial formula. To find the value of n C„ we can use either the
combinations formula from Section 5.5.2 or a calculator.
To find the probability of x successes in n trials for a binomial experiment, the only values
needed are those of n and p. These are called the parameters of the binomial probability
distribution or simply the binomial parameters. The value of q is obtained by subtracting the
value of p from 1.0. Thus, q = 1 — p.
Next we solve a binomial problem, first without using the binomial formula and then by
using the binomial formula.
Calculating the probability
using a tree diagram and the
binomial formula.
■ EXAMPLE 5-18
Five percent of all DVD players manufactured by a large electronics company are defective.
A quality control inspector randomly selects three DVD players from the production line. What
is the probability that exactly one of these three DVD players is defective?
5.6 The Binomial Probability Distribution 217
Solution Let
D = a selected DVD player is defective
G = a selected DVD player is good
As the tree diagram in Figure 5.4 shows, there are a total of eight outcomes, and three of
them contain exactly one defective DVD player. These three outcomes are
DGG, GDG, and GGD
We know that 5% of all DVD players manufactured at this company are defective. As a re-
sult, 95% of all DVD players are good. So the probability that a randomly selected DVD player
is defective is .05 and the probability that it is good is .95.
P(D) = .05 and P(G) = .95
Because the size of the population is large (note that it is a large company), the selections
can be considered to be independent. The probability of each of the three outcomes that give
exactly one defective DVD player is calculated as follows:
P{DGG) = P{D) • P{G) • P{G) = (.05)(.95)(.95) = .0451
P{GDG) = P{G) ■ P{D) • P{G) = (.95)(.05)(.95) = .0451
P{GGD) = P{G) ■ P{G) • P{D) = (.95)(.95)(.05) = .0451
Note that DGG is simply the intersection of the three events D, G, and G. In other words,
P(DGG) is the joint probability of three events: the first DVD player selected is defective, the
second is good, and the third is good. To calculate this probability, we use the multiplication
rule for independent events we learned in Chapter 4. The same is true about the probabilities
of the other two outcomes: GDG and GGD.
218 Chapter 5 Discrete Random Variables and Their Probability Distributions
Exactly one defective DVD player will be selected if DGG or GDG or GGD occurs. These
are three mutually exclusive outcomes. Therefore, from the addition rule of Chapter 4, the
probability of the union of these three outcomes is simply the sum of their individual prob-
abilities.
P(l DVD player in 3 is defective) = P(DGG or GDG or GGD)
= P{DGG) + P(GDG) + P{GGD)
= .0451 + .0451 + .0451 = .1353
Now let us use the binomial formula to compute this probability. Let us call the selection
of a defective DVD player a success and the selection of a good DVD player a. failure. The
reason we have called a defective DVD player a success is that the question refers to select-
ing exactly one defective DVD player. Then,
n = total number of trials = 3 DVD players
x = number of successes = number of defective DVD players = 1
n — x = number of failures = number of good DVD players = 3 — 1=2
p = P(success) = .05
q = P(failure) = 1 - p = .95
The probability of one success is denoted by P(x = 1) or simply by P(l). By substituting
all the values in the binomial formula, we obtain
Number of ways to Number of Number of
obtain 1 success in successes failures
3 trials —
1
r
P(x = 1) = 3 C,(.05) 1 (.95) 2 = (3)(.05)(.9025) = .1354
Probability
Probability
of success of failure
Note that the value of 3 C l in the formula either can be obtained from a calculator or can
be computed as follows:
3! _ 3-2-1 _
sCl ~ 1!(3 - 1)! ~ 1-2-1 " 3
In the above computation, 3 Ci gives the three ways to select one defective DVD player in
three selections. As listed previously, these three ways to select one defective DVD player
are DGG, GDG, and GGD. The probability .1354 is slightly different from the earlier calcu-
lation (.1353) because of rounding. ■
Calculating the probability
using the binomial formula.
■ EXAMPLE 5-19
At the Express House Delivery Service, providing high-quality service to customers is the top
priority of the management. The company guarantees a refund of all charges if a package it
is delivering does not arrive at its destination by the specified time. It is known from past data
that despite all efforts, 2% of the packages mailed through this company do not arrive at their
destinations within the specified time. Suppose a corporation mails 10 packages through
Express House Delivery Service on a certain day.
(a) Find the probability that exactly one of these 10 packages will not arrive at its desti-
nation within the specified time.
(b) Find the probability that at most one of these 10 packages will not arrive at its desti-
nation within the specified time.
5.6 The Binomial Probability Distribution 219
Solution Let us call it a success if a package does not arrive at its destination within the
specified time and a failure if it does arrive within the specified time. Then,
n = total number of packages mailed = 10
p = P(success) = .02
q = P(failure) = 1 - .02 = .98
(a) For this part,
x = number of successes = 1
n — x = number of failures = 10 — 1 =9
Substituting all values in the binomial formula, we obtain
10'
P(x = 1) = 10 C,(.02)'(.98) 9 = ^ (.02) 1 (.98) 9
= (10)(.02)(.83374776) = .1667
Thus, there is a .1667 probability that exactly one of the 10 packages mailed will not
arrive at its destination within the specified time.
(b) The probability that at most one of the 10 packages will not arrive at its destination within
the specified time is given by the sum of the probabilities of x = and x = 1 . Thus,
P(x < 1) = P(x = 0) + P(x = 1)
= 10 C (.02)°(.98) 10 + 10 C,(.02) 1 (.98) 9
= (1)(1)(.81707281) + (10)(.02)(.83374776)
= .8171 + .1667 = .9838
Thus, the probability that at most one of the 10 packages will not arrive at its desti-
nation within the specified time is .9838. H
EXAMPLE 5-20
In a Robert Half International survey of senior executives, 35% of the executives said that
good employees leave companies because they are unhappy with the management (USA TODAY,
Constructing a binomial
probability distribution
February 10, 2009). Assume that this result holds true for the current population of senior and its graph
executives. Let x denote the number in a random sample of three senior executives who hold
this opinion. Write the probability distribution of x and draw a bar graph for this probability
distribution.
Solution Let x be the number of senior executives in a sample of three who hold the said
opinion. Then, n — x is the number of senior executives who do not hold this opinion. From
the given information,
n = total senior executives in the sample = 3
p = P(a senior executive holds the said opinion) = .35
q = P(a senior executive does not hold the said opinion) = 1 — .35 = .65
The possible values that x can assume are 0, 1,2, and 3. In other words, the number of sen-
ior executives in a sample of three who hold the said opinion can be 0, 1, 2, or 3. The prob-
ability of each of these four outcomes is calculated as follows.
If x = 0, then n — x = 3. Using the binomial formula, we obtain the probability of x = as
P(x = 0) = 3 C (.35)°(.65) 3 = (1)(1)(.274625) = .2746
Note that 3 C is equal to 1 by definition, and (.35)° is equal to 1 because any number raised to
the power zero is always 1 .
220 Chapter 5 Discrete Random Variables and Their Probability Distributions
If x = 1, then n — x = 2. Using the binomial formula, we find the probability of x = 1 as
P(x = 1) = 3 C 1 (.35) 1 (.65) 2 = (3)(.35)(.4225) = .4436
Similarly, if x = 2, then n — x = 1, and if x = 3, then n — x = 0. The probabilities of x = 2
and x = 3 are, respectively,
P(x = 2) = 3 C 2 (.35) 2 (.65)' = (3)(.1225)(.65) = .2389
P{x = 3) = 3 C 3 (.35) 3 (.65) = (1)(.042875)(1) = .0429
These probabilities are written in Table 5.9. Figure 5.5 shows the bar graph for the probabil-
ity distribution of Table 5.9.
Table 5.9 Probability
Distribution
of x
x P(x)
.2746
1 .4436
2 .2389
3 .0429
12 3
5.6.3 Using the Table of Binomial Probabilities
The probabilities for a binomial experiment can also be read from Table I, the table of bino-
mial probabilities, in Appendix C. That table lists the probabilities of x for n = 1 to n = 25
and for selected values of p. Example 5-21 illustrates how to read Table I.
■ EXAMPLE 5-21
In an Accountemps survey of senior executives, 30% of the senior executives said that it is
appropriate for job candidates to ask about compensation and benefits during the first inter-
view (USA TODAY, April 13, 2009). Suppose that this result holds true for the current popu-
lation of senior executives in the United States. A random sample of six senior executives is
selected. Using Table I of Appendix C, answer the following.
(a) Find the probability that exactly three senior executives in this sample hold the said
opinion.
(b) Find the probability that at most two senior executives in this sample hold the said
opinion.
(c) Find the probability that at least three senior executives in this sample hold the said
opinion.
(d) Find the probability that one to three senior executives in this sample hold the said
opinion.
(e) Let x be the number of senior executives in this sample who hold the said opinion.
Write the probability distribution of x, and draw a bar graph for this probability
distribution.
Solution
(a) To read the required probability from Table I of Appendix C, we first determine the
values of n, x, and p. For this example,
Figure 5.5 Bar graph ot the P(x)
probability distribution of x.
0.5
0.4
0.3
0.2
0.1
JZL
Using the binomial table to
find probabilities and to
construct the probability
distribution and graph.
5.6 The Binomial Probability Distribution 221
n = number of senior executives in the sample = 6
x = number of senior executives in this sample who hold the said opinion = 3
p = P(a senior executive holds the said opinion) = .30
Then we locate n = 6 in the column labeled n in Table I of Appendix C. The relevant portion
of Table I with n = 6 is reproduced as Table 5.10. Next, we locate 3 in the column for x in
the portion of the table for n = 6 and locate p =.30 in the row for p at the top of the table.
The entry at the intersection of the row for x = 3 and the column for p =.30 gives the prob-
ability of three successes in six trials when the probability of success is .30. From Table I or
Table 5.10,
P(x = 3) = .1852
Table 5.10
Determining P(x
= 3) for n
= 6 and p
= .30
= .30
n
X
P
.05
.10
.20
[30]
.95
m
.7351
.5314
.2621
.1176
.0000
1
.2321
.3543
.3932
.3025
.0000
2
.0305
.0984
.2458
.3241
.0001
= 3 —
.0021
.0146
.0819
|.1852|< —
.0021
4
.0001
.0012
.0154
.0595
.0305
5
.0000
.0001
.0015
.0102
.2321
6
.0000
.0000
.0001
.0007
.7351
P(x = 3) = .1852
Using Table I or Table 5.10, we write Table 5.11, which can be used to answer the remaining
parts of this example.
Table 5.11 Portion of Table I for
n = 6 and p = .30
11
X
P
.30
6
.1176
1
.3025
2
.3241
3
.1852
4
.0595
5
.0102
6
.0007
(b) The event that at most two senior executives in this sample hold the said opinion will
occur if x is equal to 0, 1, or 2. From Table I of Appendix C or Table 5.11, the re-
quired probability is
P(at most 2) = P(Q or 1 or 2) = P(x = 0) + P(x = 1) + P(x = 2)
= .1176 + .3025 + .3241 = .7442
(c) The probability that at least three senior executives in this sample hold the said opinion
is given by the sum of the probabilities of 3, 4, 5, or 6. Using Table I of Appendix C
or Table 5.11, we obtain
222 Chapter 5 Discrete Random Variables and Their Probability Distributions
P(at least 3) = P(3 or 4 or 5 or 6)
= P(x = 3) + P(x = 4) + P(x = 5) + P(x = 6)
= .1852 + .0595 + .0102 + .0007 = .2556
(d) The probability that one to three senior executives in this sample hold the said opinion
is given by the sum of the probabilities of x = 1, 2, or 3. Using Table I of Appendix
C or Table 5.11, we obtain
P(l to 3) = P(x = 1) + P(x = 2) + P(x = 3)
= .3025 + .3241 + .1852 = .8118
(e) Using Table I of Appendix C or Table 5.11, we list the probability distribution of x for
n = 6 and p =.30 in Table 5.12. Figure 5.6 shows the bar graph of the probability dis-
tribution of x.
Table 5.12
Probability Distribution of
P{x)
x for n = 6 and p = .30
.35
X
P(x)
.30
.1176
.25
1
.3025
.20
2
.3241
.15
3
.1852
.10
4
.0595
.05
5
.0102
6
.0007
1 2 3 4 5 6 x
Figure 5.6 Bar graph for the probability distribution of x.
5.6.4 Probability of Success and the Shape
of the Binomial Distribution
For any number of trials n:
1. The binomial probability distribution is symmetric if p = .50.
2. The binomial probability distribution is skewed to the right if p is less than .50.
3. The binomial probability distribution is skewed to the left if p is greater than .50.
These three cases are illustrated next with examples and graphs.
1. Let n = 4 and p = .50. Using Table I of Appendix C, we have written the probability dis-
tribution of x in Table 5.13 and plotted it in Figure 5.7. As we can observe from Table 5.13
and Figure 5.7, the probability distribution of x is symmetric.
Table 5.13
Probability
P(x)
Distribution
of x for n = 4
.40
and p — .50
x
P(x)
.30
.0625
.20
1
.2500
2
.3750
.10
3
.2500
4
.0625
1
Figure 5.7 Bar graph for the probability distribution
of Table 5.13.
5.6 The Binomial Probability Distribution 223
2. Let n = A and p = .30 (which is less than .50). Table 5.14, which is written by using
Table I of Appendix C, and the graph of the probability distribution in Figure 5.8 show
that the probability distribution of x for n = 4 and p = .30 is skewed to the right.
laoie 3.14
Probability
Distribution
fKX)
of x for n = 4
.40
and p = .30
X
P(x)
.30
.2401
.20
1
.4116
2
.2646
.10
3
.0756
4
.0081
Figure 5.8 Bar graph for the probability distribution
of Table 5.14.
3. Let n = 4 and p = .80 (which is greater than .50). Table 5.15, which is written by using
Table I of Appendix C, and the graph of the probability distribution in Figure 5.9 show that
the probability distribution of x for n = 4 and p = .80 is skewed to the left.
Table 5.15
Probability
Distribution
P(x)
of x for n = 4
.40
and p = .80
X
P(x)
.30
.0016
.20
1
.0256
2
.1536
.10
3
.4096
4
.4096
1
Figure 5.9 Bar graph for the probability distribution
of Table 5.15.
5.6.5 Mean and Standard Deviation
of the Binomial Distribution
Sections 5.3 and 5.4 explained how to compute the mean and standard deviation, respec-
tively, for a probability distribution of a discrete random variable. When a discrete random
variable has a binomial distribution, the formulas learned in Sections 5.3 and 5.4 could still
be used to compute its mean and standard deviation. However, it is simpler and more con-
venient to use the following formulas to find the mean and standard deviation in such cases.
Mean and Standard Deviation of a Binomial Distribution The mean and standard deviation of a
binomial distribution are, respectively,
jx = np and a = \fnpq
where n is the total number of trials, p is the probability of success, and q is the probability of failure.
224 Chapter 5 Discrete Random Variables and Their Probability Distributions
Example 5-22 describes the calculation of the mean and standard deviation of a binomial
distribution.
■ EXAMPLE 5-22
According to a Harris Interactive survey conducted for World Vision and released in February
2009, 56% of teens in the United States volunteer time for charitable causes. Assume that this
result is true for the current population of U.S. teens. A sample of 60 teens is selected. Let x
be the number of teens in this sample who volunteer time for charitable causes. Find the mean
and standard deviation of the probability distribution of x.
Solution This is a binomial experiment with a total of 60 trials. Each trial has two out-
comes: (1) the selected teen volunteers time for charitable causes or (2) the selected teen does
not volunteer time for charitable causes. The probabilities p and q for these two outcomes
are .56 and .44, respectively. Thus,
n = 60, p = .56, and q = .44
Using the formulas for the mean and standard deviation of the binomial distribution, we
obtain
/x — up — 60 (.56) = 33.60
<r = Vnpq = V(60)(.56)(.44) = 3.845
Thus, the mean of the probability distribution of x is 33.60 and the standard deviation is
3.845. The value of the mean is what we expect to obtain, on average, per repetition of the
experiment. In this example, if we select many samples of 60 teens each, we expect that each
sample will contain an average of 33.60 teens, with a standard deviation of 3.845, who vol-
unteer time for charitable causes.
Calculating the mean and
standard deviation of a
binomial random variable.
EXERCISES
CONCEPTS AND PROCEDURES
5.49 Briefly explain the following.
a. A binomial experiment b. A trial c. A binomial random variable
5.50 What are the parameters of the binomial probability distribution, and what do they mean?
5.51 Which of the following are binomial experiments? Explain why.
a. Rolling a die many times and observing the number of spots
b. Rolling a die many times and observing whether the number obtained is even or odd
c. Selecting a few voters from a very large population of voters and observing whether or not each
of them favors a certain proposition in an election when 54% of all voters are known to be in fa-
vor of this proposition.
5.52 Which of the following are binomial experiments? Explain why.
a. Drawing 3 balls with replacement from a box that contains 10 balls, 6 of which are red and 4 are
blue, and observing the colors of the drawn balls
b. Drawing 3 balls without replacement from a box that contains 10 balls, 6 of which are red and 4
are blue, and observing the colors of the drawn balls
c. Selecting a few households from New York City and observing whether or not they own stocks
when it is known that 28% of all households in New York City own stocks
5.53 Let x be a discrete random variable that possesses a binomial distribution. Using the binomial for-
mula, find the following probabilities.
a. P(x = 5) for n = 8 and p = .70
b. P(x = 3) for n = 4 and p = .40
c. P{x = 2) for n = 6 and p = .30
Verify your answers by using Table I of Appendix C.
5.6 The Binomial Probability Distribution 225
5.54 Let x be a discrete random variable that possesses a binomial distribution. Using the binomial for-
mula, find the following probabilities.
a. P(x = 0) for n = 5 and p = .05
b. P(x = 4) for n = 7 and p = .90
c. P(x = 7) for n = 10 and p = .60
Verify your answers by using Table I of Appendix C.
5.55 Let x be a discrete random variable that possesses a binomial distribution.
a. Using Table I of Appendix C, write the probability distribution of x for n = 7 and p = .30 and
graph it.
b. What are the mean and standard deviation of the probability distribution developed in part a?
5.56 Let x be a discrete random variable that possesses a binomial distribution.
a. Using Table I of Appendix C, write the probability distribution of x for n = 5 and p = .80 and
graph it.
b. What are the mean and standard deviation of the probability distribution developed in part a?
5.57 The binomial probability distribution is symmetric for p = .50, skewed to the right for p < .50, and
skewed to the left for p > .50. Illustrate each of these three cases by writing a probability distribution table
and drawing a graph. Choose any values of n and p and use the table of binomial probabilities (Table I of
Appendix C) to write the probability distribution tables.
■ APPLICATIONS
5.58 According to a Harris Interactive poll, 52% of American college graduates have Facebook accounts
(http://www.harrisinteractive.com/harris_poll/pubs/Harris_Poll_2009_04_16.pdf). Suppose that this result
is true for the current population of American college graduates.
a. Let x be a binomial random variable that denotes the number of American college graduates in
a random sample of 15 who have Facebook accounts. What are the possible values that x can
assume?
b. Find the probability that exactly 9 American college graduates in a sample of 15 have Face-
book accounts.
5.59 According to a National Public Radio poll, 46% of American school principals believe that students
pay little or no attention to sex education provided in schools (http://www.npr.org/programs/morning/
features/2004/jan/kaiserpoll/principalsfinal.pdf). Suppose that this result is true for the current population
of American school principals.
a. Let x be a binomial random variable denoting the number of American school principals in a ran-
dom sample of 1 1 who believe that students pay little or no attention to sex education taught in
schools. What are the possible values that x can assume?
b. Find the probability that in a random sample of 1 1 American school principals, exactly 3 believe
that students pay little or no attention to sex education taught in schools.
5.60 In a 2009 poll of adults 18 years and older, (BBMG Conscious Consumer Report) about half of
them said that despite tough economic times, they are willing to pay more for products that have social
and environmental benefits. Suppose that 50% of all such adults currently hold this view. Suppose that
a random sample of 20 such adults is selected. Use the binomial probabilities table (Table I of Appen-
dix C) or technology to find the probability that the number of adults in this sample who hold this
opinion is
a. at most 7 b. at least 13 c. 12 to 15
5.61 According to a March 25, 2007 Pittsburgh Post-Gazette article, 30% to 40% of U.S. taxpayers cheat
on their returns. Suppose that 30% of all current U.S. taxpayers cheat on their returns. Use the binomial
probabilities table (Table I of Appendix C) or technology to find the probability that the number of U.S.
taxpayers in a random sample of 14 who cheat on their taxes is
a. at least 8 b. at most 3 c. 3 to 7
5.62 Magnetic resonance imaging (MRI) is a process that produces internal body images using a strong
magnetic field. Some patients become claustrophobic and require sedation because they are required to lie
within a small, enclosed space during the MRI test. Suppose that 20% of all patients undergoing MRI test-
ing require sedation due to claustrophobia. If five patients are selected at random, find the probability that
the number of patients in these five who require sedation is
a. exactly 2 b. none c. exactly 4
226
Chapter 5 Discrete Random Variables and Their Probability Distributions
5.63 According to an October 27, 2006 article in Newsweek, 65% of Americans said that they take ex-
pired medicines. Suppose that this result is true of the current population of Americans. Find the proba-
bility that the number of Americans in a random sample of 22 who take expired medicines is
a. exactly 17 b. none c. exactly 9
5.64 According to Case Study 4-2 in Chapter 4, the probability that a baseball player will have no hits
in 10 trips to the plate is .0563, given that this player has a batting average of .250. Using the binomial
formula, show that this probability is indeed .0563.
5.65 A professional basketball player makes 85% of the free throws he tries. Assuming this percentage
will hold true for future attempts, find the probability that in the next eight tries, the number of free throws
he will make is
a. exactly 8 b. exactly 5
5.66 According to a 2008 Pew Research Center survey of adult men and women, close to 70% of these
adults said that men and women possess equal traits for being leaders. Suppose 70% of the current pop-
ulation of adults holds this view.
a. Using the binomial formula, find the probability that in a sample of 16 adults, the number who
will hold this view is
i. exactly 13 ii. exactly 16
b. Use the binomial probabilities table (Table I of Appendix C) or technology to find the probabil-
ity that the number of adults in this sample who will hold this view is
i. at least 11 ii. at most 8 iii. 9 to 12
5.67 An office supply company conducted a survey before marketing a new paper shredder designed
for home use. In the survey, 80% of the people who used the shredder were satisfied with it. Because
of this high acceptance rate, the company decided to market the new shredder. Assume that 80% of
all people who will use it will be satisfied. On a certain day, seven customers bought this shredder.
a. Let x denote the number of customers in this sample of seven who will be satisfied with this
shredder. Using the binomial probabilities table (Table I, Appendix C), obtain the probability
distribution of x and draw a graph of the probability distribution. Find the mean and standard
deviation of x.
b. Using the probability distribution of part a, find the probability that exactly four of the seven cus-
tomers will be satisfied.
5.68 Johnson Electronics makes calculators. Consumer satisfaction is one of the top priorities of the com-
pany's management. The company guarantees a refund or a replacement for any calculator that malfunc-
tions within 2 years from the date of purchase. It is known from past data that despite all efforts, 5% of
the calculators manufactured by the company malfunction within a 2-year period. The company mailed a
package of 10 randomly selected calculators to a store.
a. Let x denote the number of calculators in this package of 10 that will be returned for refund or
replacement within a 2-year period. Using the binomial probabilities table, obtain the probability
distribution of x and draw a graph of the probability distribution. Determine the mean and stan-
dard deviation of x.
b. Using the probability distribution of part a, find the probability that exactly 2 of the 10 calcula-
tors will be returned for refund or replacement within a 2-year period.
5.69 A fast food chain store conducted a taste survey before marketing a new hamburger. The results of
the survey showed that 70% of the people who tried this hamburger liked it. Encouraged by this result,
the company decided to market the new hamburger. Assume that 70% of all people like this hamburger.
On a certain day, eight customers bought it for the first time.
a. Let x denote the number of customers in this sample of eight who will like this hamburger. Us-
ing the binomial probabilities table, obtain the probability distribution of x and draw a graph of
the probability distribution. Determine the mean and standard deviation of x.
b. Using the probability distribution of part a, find the probability that exactly three of the eight cus-
tomers will like this hamburger.
5.7 The Hypergeometric Probability Distribution
In Section 5.6, we learned that one of the conditions required to apply the binomial probabil-
ity distribution is that the trials are independent, so that the probabilities of the two outcomes
(success and failure) remain constant. If the trials are not independent, we cannot apply the
5.7 The Hypergeometric Probability Distribution 227
binomial probability distribution to find the probability of x successes in n trials. In such cases
we replace the binomial by the hypergeometric probability distribution. Such a case occurs
when a sample is drawn without replacement from a finite population.
As an example, suppose 20% of all auto parts manufactured at a company are defective.
Four auto parts are selected at random. What is the probability that three of these four parts are
good? Note that we are to find the probability that three of the four auto parts are good and one
is defective. In this case, the population is very large and the probability of the first, second,
third, and fourth auto parts being defective remains the same at .20. Similarly, the probability
of any of the parts being good remains unchanged at .80. Consequently, we will apply the bi-
nomial probability distribution to find the probability of three good parts in four.
Now suppose this company shipped 25 auto parts to a dealer. Later, it finds out that 5 of
those parts were defective. By the time the company manager contacts the dealer, 4 auto parts
from that shipment have already been sold. What is the probability that 3 of those 4 parts were
good parts and 1 was defective? Here, because the 4 parts were selected without replacement
from a small population, the probability of a part being good changes from the first selection
to the second selection, to the third selection, and to the fourth selection. In this case we can-
not apply the binomial probability distribution. In such instances, we use the hypergeometric
probability distribution to find the required probability.
Hypergeometric Probability Distribution
Let
N = total number of elements in the population
r = number of successes in the population
N — r = number of failures in the population
n = number of trials (sample size)
x = number of successes in n trials
n — x = number of failures in n trials
The probability of x successes in n trials is given by
Examples 5-23 and 5-24 provide applications of the hypergeometric probability distribution.
■ EXAMPLE 5-23
Brown Manufacturing makes auto parts that are sold to auto dealers. Last week the company
shipped 25 auto parts to a dealer. Later, it found out that 5 of those parts were defective. By
the time the company manager contacted the dealer, 4 auto parts from that shipment had
already been sold. What is the probability that 3 of those 4 parts were good parts and 1 was
defective?
Calculating probability by
using hypergeometric
distribution formula.
Solution Let a good part be called a success and a defective part be called a failure. From
the given information,
N = total number of elements (auto parts) in the population = 25
r = number of successes (good parts) in the population = 20
N — r = number of failures (defective parts) in the population = 5
228 Chapter 5 Discrete Random Variables and Their Probability Distributions
n = number of trials (sample size) = 4
x = number of successes in four trials = 3
n — x = number of failures in four trials = 1
Using the hypergeometric formula, we calculate the required probability as follows:
20! 5!
.(\ \ X): ,. 2C A 5C1 3!(20 - 3)! ' 11(5 - 1)!
Pyx = 3) = = =
jvC„ 25 C4 25!
4!(25 - 4)!
(1140) (5)
12,650
.4506
Thus, the probability that 3 of the 4 parts sold are good and 1 is defective is .4506.
In the above calculations, the values of combinations can either be calculated using the for-
mula learned in Section 5.5.2 (as done here) or by using a calculator. I
Calculating probability by
using hypergeometric
distribution formula.
■ EXAMPLE 5-24
Dawn Corporation has 12 employees who hold managerial positions. Of them, 7 are female
and 5 are male. The company is planning to send 3 of these 12 managers to a conference. If
3 managers are randomly selected out of 12,
(a) find the probability that all 3 of them are female
(b) find the probability that at most 1 of them is a female
Solution Let the selection of a female be called a success and the selection of a male be
called a failure.
(a) From the given information,
N = total number of managers in the population
(b)
12
N
r
r
n
x
x
number of successes (females) in the population = 7
number of failures (males) in the population = 5
number of selections (sample size) = 3
number of successes (females) in three selections = 3
number of failures (males) in three selections =
Using the hypergeometric formula, we calculate the required probability as follows:
r C xN - r C„- K 7 C 35 C (35)(1)
P(x = 3)
.1591
N C„ U C 3 220
Thus, the probability that all 3 of the managers selected are female is .1591.
The probability that at most 1 of them is a female is given by the sum of the proba-
bilities that either none or 1 of the selected managers is a female.
To find the probability that none of the selected managers is a female, we use
N = total number of managers in the population =12
r = number of successes (females) in the population = 7
N — r = number of failures (males) in the population = 5
n = number of selections (sample size) = 3
x = number of successes (females) in three selections =
n — x = number of failures (males) in three selections = 3
5.7 The Hypergeometric Probability Distribution
229
Using the hypergeometric formula, we calculate the required probability as follows:
p {x = o) = = ^1 = MM = , 0455
N C n 12 C 3 220
To find the probability that 1 of the selected managers is a female, we use
N = total number of managers in the population =12
r = number of successes (females) in the population = 7
N — r = number of failures (males) in the population = 5
n = number of selections (sample size) = 3
x = number of successes (females) in three selections = 1
n — x = number of failures (males) in three selections = 2
Using the hypergeometric formula, we obtain the required probability as follows:
. r C xN _ r C n _ x 7 C 15 C 2 (7)(10)
P(x = 1) = = = = .3182
N C„ 12 C 3 220
The probability that at most 1 of the 3 managers selected is a female is
P(x < 1) = P(x = 0) + P(x = 1) = .0455 + .3182 = .3637 ■
EXERCISES
CONCEPTS AND PROCEDURES
5.70 Explain the hypergeometric probability distribution. Under what conditions is this probability distri-
bution applied to find the probability of a discrete random variable xl Give one example of the applica-
tion of the hypergeometric probability distribution.
5.71 Let N = 8, r = 3, and n = 4. Using the hypergeometric probability distribution formula, find
a. P(x = 2) b. P(x = 0) c. P(x < 1)
5.72 Let N = 14, r = 6, and n = 5. Using the hypergeometric probability distribution formula, find
a. P(x = 4) b. P(x = 5) c. P(x < 1)
5.73 Let N = 11, r = 4, and n = 4. Using the hypergeometric probability distribution formula, find
a. P(x = 2) b. P(x = 4) c. P(x < 1)
5.74 Let N = 16, r = 10, and n = 5. Using the hypergeometric probability distribution formula, find
a. P(x = 5) b. P(x = 0) c. P(x < 1)
■ APPLICATIONS
5.75 An Internal Revenue Service inspector is to select 3 corporations from a list of 15 for tax audit purposes. Of
the 15 corporations, 6 earned profits and 9 incurred losses during the year for which the tax returns are to be
audited. If the IRS inspector decides to select 3 corporations randomly, find the probability that the number
of corporations in these 3 that incurred losses during the year for which the tax returns are to be audited is
a. exactly 2 b. none c. at most 1
5.76 Six jurors are to be selected from a pool of 20 potential candidates to hear a civil case involving a
lawsuit between two families. Unknown to the judge or any of the attorneys, 4 of the 20 prospective ju-
rors are potentially prejudiced by being acquainted with one or more of the litigants. They will not dis-
close this during the jury selection process. If 6 jurors are selected at random from this group of 20, find
the probability that the number of potentially prejudiced jurors among the 6 selected jurors is
a. exactly 1 b. none c. at most 2
5.77 A really bad carton of 18 eggs contains 7 spoiled eggs. An unsuspecting chef picks 4 eggs at ran-
dom for his "Mega-Omelet Surprise." Find the probability that the number of unspoiled eggs among the
4 selected is
a. exactly 4 b. 2 or fewer c. more than 1
230
Chapter 5 Discrete Random Variables and Their Probability Distributions
5.78 Bender Electronics buys keyboards for its computers from another company. The keyboards are re-
ceived in shipments of 100 boxes, each box containing 20 keyboards. The quality control department at
Bender Electronics first randomly selects one box from each shipment and then randomly selects 5 key-
boards from that box. The shipment is accepted if not more than 1 of the 5 keyboards is defective. The
quality control inspector at Bender Electronics selected a box from a recently received shipment of key-
boards. Unknown to the inspector, this box contains 6 defective keyboards.
a. What is the probability that this shipment will be accepted?
b. What is the probability that this shipment will not be accepted?
5.8 The Poisson Probability Distribution
The Poisson probability distribution, named after the French mathematician Simeon-Denis.
Poisson, is another important probability distribution of a discrete random variable that has a
large number of applications. Suppose a washing machine in a laundromat breaks down an
average of three times a month. We may want to find the probability of exactly two breakdowns
during the next month. This is an example of a Poisson probability distribution problem. Each
breakdown is called an occurrence in Poisson probability distribution terminology. The Poisson
probability distribution is applied to experiments with random and independent occurrences.
The occurrences are random in the sense that they do not follow any pattern, and, hence, they
are unpredictable. Independence of occurrences means that one occurrence (or nonoccurrence)
of an event does not influence the successive occurrences or nonoccurrences of that event. The
occurrences are always considered with respect to an interval. In the example of the washing ma-
chine, the interval is one month. The interval may be a time interval, a space interval, or a vol-
ume interval. The actual number of occurrences within an interval is random and independent.
If the average number of occurrences for a given interval is known, then by using the Poisson
probability distribution, we can compute the probability of a certain number of occurrences, x,
in that interval. Note that the number of actual occurrences in an interval is denoted by x.
Conditions to Apply the Poisson Probability Distribution The following three conditions must
be satisfied to apply the Poisson probability distribution.
1. x is a discrete random variable.
2. The occurrences are random.
3. The occurrences are independent.
The following are three examples of discrete random variables for which the occurrences
are random and independent. Hence, these are examples to which the Poisson probability dis-
tribution can be applied.
1. Consider the number of telemarketing phone calls received by a household during a given day.
In this example, the receiving of a telemarketing phone call by a household is called an occur-
rence, the interval is one day (an interval of time), and the occurrences are random (that is, there
is no specified time for such a phone call to come in) and discrete. The total number of tele-
marketing phone calls received by a household during a given day may be 0, 1, 2, 3, 4, and so
forth. The independence of occurrences in this example means that the telemarketing phone
calls are received individually and none of two (or more) of these phone calls are related.
2. Consider the number of defective items in the next 100 items manufactured on a machine. In
this case, the interval is a volume interval (100 items). The occurrences (number of defective
items) are random and discrete because there may be 0, 1, 2, 3, 100 defective items in 100
items. We can assume the occurrence of defective items to be independent of one another.
3. Consider the number of defects in a 5-foot-long iron rod. The interval, in this example, is
a space interval (5 feet). The occurrences (defects) are random because there may be any
number of defects in a 5-foot iron rod. We can assume that these defects are independent
of one another.
5.8 The Poisson Probability Distribution 231
The following examples also qualify for the application of the Poisson probability distribution.
1. The number of accidents that occur on a given highway during a 1-week period
2. The number of customers entering a grocery store during a 1-hour interval
3. The number of television sets sold at a department store during a given week
In contrast, consider the arrival of patients at a physician's office. These arrivals are non-
random if the patients have to make appointments to see the doctor. The arrival of commercial
airplanes at an airport is nonrandom because all planes are scheduled to arrive at certain times,
and airport authorities know the exact number of arrivals for any period (although this number
may change slightly because of late or early arrivals and cancellations). The Poisson probabil-
ity distribution cannot be applied to these examples.
In the Poisson probability distribution terminology, the average number of occurrences in
an interval is denoted by X (Greek letter lambda). The actual number of occurrences in that in-
terval is denoted by x. Then, using the Poisson probability distribution, we find the probability
of x occurrences during an interval given that the mean occurrences during that interval are X.
Poisson Probability Distribution Formula According to the Poisson probability distribution, the
probability of x occurrences in an interval is
x!
where X (pronounced lambda) is the mean number of occurrences in that interval and the value
of e is approximately 2.71828.
The mean number of occurrences in an interval, denoted by X, is called the parameter of the
Poisson probability distribution or the Poisson parameter. As is obvious from the Poisson prob-
ability distribution formula, we need to know only the value of X to compute the probability of
any given value of x. We can read the value of e ~ k for a given X from Table II of Appendix C.
Examples 5-25 through 5-27 illustrate the use of the Poisson probability distribution formula.
■ EXAMPLE 5-25
On average, a household receives 9.5 telemarketing phone calls per week. Using the Poisson
Using the Poisson formula:
probability distribution formula, find the probability that a randomly selected household , .„ ,
r j J x equals a specific value.
receives exactly 6 telemarketing phone calls during a given week.
Solution Let A be the mean number of telemarketing phone calls received by a household
per week. Then, A = 9.5. Let x be the number of telemarketing phone calls received by a
household during a given week. We are to find the probability of x = 6. Substituting all the
values in the Poisson formula, we obtain
X x e~ x (9.5)V 5 (735,09 1.8906)(.00007485)
p(x = 6) = ^r = ^^ = 720 = 0764
To do these calculations, we can find the value of 6! either by using the factorial key on a
calculator or by multiplying all integers from 1 to 6, and we can find the value of e by us
ing the e x key on a calculator or from Table II in Appendix C.
EXAMPLE 5-26
A washing machine in a laundromat breaks down an average of three times per month. Using Calculating probabilities
the Poisson probability distribution formula, find the probability that during the next month using the Poisson formula
this machine will have
(a) exactly two breakdowns
(b) at most one breakdown
232 Chapter 5 Discrete Random Variables and Their Probability Distributions
Solution Let A be the mean number of breakdowns per month, and let x be the actual num-
ber of breakdowns observed during the next month for this machine. Then,
A = 3
(a) The probability that exactly two breakdowns will be observed during the next month is
A'e A (3) 2 e~ 3 (9)(.04978707)
P(x = 2) = = — = — = .2240
v ' x\ 2! 2
(b) The probability that at most one breakdown will be observed during the next month
is given by the sum of the probabilities of zero and one breakdown. Thus,
f(at most 1 breakdown) = P(0 or 1 breakdown) = P(x = 0) + P(x = 1)
(3)V 3 (3)' e - 3
+
0! 1!
(1)(. 04978707) (3)(.04978707)
1 + 1
= .0498 + .1494 = .1992 ■
Remember ► One important point about the Poisson probability distribution is that the intervals for A and x
must be equal. If they are not, the mean A should be redefined to make them equal. Example
5-27 illustrates this point.
■ EXAMPLE 5-27
Cynthia's Mail Order Company provides free examination of its products for 7 days. If not
Calculating a probability completely satisfied, a customer can return the product within that period and get a full re-
fund. According to past records of the company, an average of 2 of every 10 products sold by
this company are returned for a refund. Using the Poisson probability distribution formula,
find the probability that exactly 6 of the 40 products sold by this company on a given day will
be returned for a refund.
Solution Let x denote the number of products in 40 that will be returned for a refund. We
are to find P(x = 6). The given mean is defined per 10 products, but x is defined for 40 prod-
ucts. As a result, we should first find the mean for 40 products. Because, on average, 2 out of
10 products are returned, the mean number of products returned out of 40 will be 8. Thus,
X = 8. Substituting x = 6 and X. = 8 in the Poisson probability distribution formula, we obtain
A' e - A (8)V 8 (262, 144)(. 00033546)
P ( x = 6 ) = ^— = y -2 = ^ * ; = .1221
y ' x\ 6! 720
Thus, the probability is .1221 that exactly 6 products out of 40 sold on a given day will be
returned. I
Note that Example 5-27 is actually a binomial problem with p = 2/10 = .20, n = 40, and
x = 6. In other words, the probability of success (that is, the probability that a product is re-
turned) is .20 and the number of trials (products sold) is 40. We are to find the probability of
six successes (returns). However, we used the Poisson distribution to solve this problem. This
is referred to as using the Poisson distribution as an approximation to the binomial distribution.
We can also use the binomial distribution to find this probability as follows:
40'
P(x = 6) = 40 C 6 (.20) 6 (.80) 34 = 6!(4Q 1 6) , (-20) 6 (-80) 34
= (3,838,380)(.000064)(.00050706) = .1246
Thus the probability P(x = 6) is .1246 when we use the binomial distribution.
Fortune magazine used to publish a column titled Ask Mr. Statistics, which contained questions and an-
swers to statistical problems. The following excerpts are reprinted from one such column.
Dear Oddgiver: I am in the seafood distribution business and find myself endlessly wrangling with
supermarkets about appropriate order sizes, especially with high-end tidbit products like our matjes
herring in superspiced wine, which we let them have for $4.25, and still they take only a half-dozen
jars, thereby running the risk of getting sold out early in the week and causing the better class of
customers to storm out empty-handed. How do I get them to realize that lowballing on inventories
is usually bad business, also to at least try a few jars of our pickled crappie balls?
-HEADED FOR A BREAKDOWN
ASK MR.
STATISTICS
Dear Picklehead: The science of statistics has much to offer people puzzled by seafood inventory prob-
lems. Your salvation lies in the Poisson distribution, "poisson" being French for fish and, of arguably
greater relevance, the surname of a 19th-century French probabilist.
Simeon Poisson's contribution was to develop a method for calculating the likelihood that a spec-
ified number of successes will occur given that (a) the probability of success on any one trial is very
low but (b) the number of trials is very high. A real world example often mentioned in the literature
concerns the distribution of Prussian cavalry deaths from getting kicked by horses in the period 1875-94.
As you would expect of Teutons, the Prussian military kept meticulous records on horse-kick
deaths in each of its army corps, and the data are neatly summarized in a 1963 book called Lady
Luck, by the late Warren Weaver. There were a total of 196 kicking deaths— these being the, er,
"successes." The "trials" were each army corps' observations on the number of kicking deaths sus-
tained in the year. So with 14 army corps and data for 20 years, there were 280 trials. We shall not
detain you with the Poisson formula, but it predicts, for example, that there will be 34.1 instances
of a corps' having exactly two deaths in a year. In fact, there were 32 such cases. Pretty good, eh?
Back to seafood. The Poisson calculation is appropriate to your case, since the likelihood of any one
customer's buying your overspiced herring is extremely small, but the number of trials— i.e., customers in
the store during a typical week— is very large. Let us say that one customer in 1,000 deigns to buy the
herring, and 6,000 customers visit the store in a week. So six jars are sold in an average week.
But the store manager doesn't care about average weeks. What he's worried about is having too
much or not enough. He needs to know the probabilities assigned to different sales levels. Our Poisson
distribution shows the following morning line: The chance of fewer than three sales— only 6.2%. Of four
to six sales: 45.5%. Chances of losing some sales if the store elects to start the week with six jars be-
cause that happens to be the average: 39.4%. If the store wants to be 90% sure of not losing sales, it
needs to start with nine jars.
There is no known solution to the problem of pickled crappie balls.
Source: Daniel Seligman, "Ask
. , , , , . Mr. Statistics," Fortune, March 7, 1 994.
Quiz: Using the Poisson probability distribution, calculate the probabilities mentioned at the end of this copyright© 1994 Time Inc Reprinted
case Study. with permission. All rights reserved.
As we can observe, simplifying the above calculations for the binomial formula is quite
complicated when n is large. It is much easier to solve this problem using the Poisson proba-
bility distribution. As a general rule, if it is a binomial problem with n > 25 but /j, < 25, then
we can use the Poisson probability distribution as an approximation to the binomial distribu-
tion. However, if n > 25 and jx > 25, we prefer to use the normal distribution as an approxi-
mation to the binomial. The latter case will be discussed in Chapter 6.
Case Study 5-3 presents applications of the binomial and Poisson probability distributions.
5.8.1 Using the Table of Poisson Probabilities
The probabilities for a Poisson distribution can also be read from Table III, in Appendix C, the
table of Poisson probabilities. The following example describes how to read that table.
■ EXAMPLE 5-28
On average, two new accounts are opened per day at an Imperial Savings Bank branch. Using
Table III of Appendix C, find the probability that on a given day the number of new accounts Us ' n S the table of Poisson
opened at this bank will be
(a) exactly 6 (b) at most 3 (c) at least 7
233
234 Chapter 5 Discrete Random Variables and Their Probability Distributions
Solution Let
A = mean number of new accounts opened per day at this bank
x = number of new accounts opened at this bank on a given day
(a) The values of A and x are
A = 2 and x = 6
In Table III of Appendix C, we first locate the column that corresponds to A = 2. In
this column, we then read the value for x = 6. The relevant portion of that table is
shown here as Table 5.16. The probability that exactly 6 new accounts will be opened
on a given day is .0120. Therefore,
P(x = 6) = .0120
Table 5.16
Portion of Table III for A = 2.0
A
X
1.1 1.2
[Ml <
.1353
1
.2707
2
.2707
3
.1804
4
.0902
5
.0361
i
|.0120|
7
.0034
8
.0009
9
.0002
Actually, Table 5. 16 gives the probability distribution of x for A = 2.0. Note that the
sum of the 10 probabilities given in Table 5.16 is .9999 and not 1.0. This is so for two
reasons. First, these probabilities are rounded to four decimal places. Second, on a given
day more than 9 new accounts might be opened at this bank. However, the probabili-
ties of 10, 11, 12, . . . new accounts are very small, and they are not listed in the table.
(b) The probability that at most three new accounts are opened on a given day is obtained
by adding the probabilities of 0, 1, 2, and 3 new accounts. Thus, using Table III of
Appendix C or Table 5.16, we obtain
P(at most 3) = P(x = 0) + P(x = 1) + P(x = 2) + P(x = 3)
= .1353 + .2707 + .2707 + .1804 = .8571
(c) The probability that at least 7 new accounts are opened on a given day is obtained by
adding the probabilities of 7, 8, and 9 new accounts. Note that 9 is the last value of x
for A = 2.0 in Table III of Appendix C or Table 5.16. Hence, 9 is the last value of x
whose probability is included in the sum. However, this does not mean that on a given
day more than 9 new accounts cannot be opened. It simply means that the probabil-
ity of 10 or more accounts is close to zero. Thus,
P(at least 7) = P(x = 7) + P(x = 8) + P(x = 9)
= .0034 + .0009 + .0002 = .0045
Source: Census Bureau Bv Shannon Reillv and Karl celles. USA today
The above chart shows that, on average, one child is born per seven seconds and one person dies per
12 seconds in the United States. The findings are based on Census Bureau data. If we assume that the
births and deaths follow the Poisson probability distribution, we can find the probability of any number of
births and deaths for a given time interval. Note that there is one birth, on average, per 7 seconds. If x is
the actual number of births during any 7-second interval, then x can assume any (integer) value such as 0,
1, 2, 3, ... . The same is true for the number of deaths per 12-second interval.
Let x be the number of (actual) births in any given seven-second interval. Then, x is a discrete ran-
dom variable that can assume any of the values 0, 1, 2, 3, . . ., and so on. We can find the probability of
any of these values of x either by using the Poisson formula or by using Table III of Appendix C. For exam-
ple, from Table III, P(x = 0) is .3679 for X = 1, which means that the probability of no birth in a given
interval of seven seconds is .3679 when X = 1 . Using that table, we can construct the probability distribution
for x.
Similarly, we can find the probability of any number of deaths in an interval of 12 seconds using Table
III when, on average, there is one death per 12 seconds.
Source: USA TODAY, December 1,
2004. Copyright © 2004, USA TODAY.
Chart reproduced with permission.
■ EXAMPLE 5-29
An auto salesperson sells an average of .9 car per day. Let x be the number of cars sold by
this salesperson on any given day. Using the Poisson probability distribution table, write the Constructing a Poisson
..... ,. ., . r „ . r , , . ... .. ., . probability distribution
probability distribution of x. Draw a graph of the probability distribution. , ...
ana graphing it.
Solution Let A be the mean number of cars sold per day by this salesperson. Hence,
A = .9. Using the portion of Table III of Appendix C that corresponds to A = .9, we write the
probability distribution of x in Table 5.17. Figure 5.10 shows the bar graph for the probabil-
ity distribution of Table 5.17.
235
236 Chapter 5 Discrete Random Variables and Their Probability Distributions
Table 5.17 Probability P(x)
Distribution
.50
ofx for A = .9
.4066 .30
1 .3659
.20
2 .1647 . .
3 .0494 .10
4 .0111 | 1
5 .0020 1 2 3 4 5 6 ~~ t
^ -0003 Figure 5. TO Bar graph for the probability distribution of
Table 5.17.
Note that 6 is the largest value of x for A = .9 listed in Table III for which the probability is
greater than zero. However, this does not mean that this salesperson cannot sell more than six cars
on a given day. What this means is that the probability of selling seven or more cars is very small.
Actually, the probability of x = 7 for A = .9 calculated by using the Poisson formula is .000039.
When rounded to four decimal places, this probability is .0000, as listed in Table III. I
5.8.2 Mean and Standard Deviation of the
Poisson Probability Distribution
For the Poisson probability distribution, the mean and variance both are equal to A, and the stan-
dard deviation is equal to VA. That is, for the Poisson probability distribution,
H = A, a 2 = A, and a = VX
For Example 5-29, A = .9. Therefore, for the probability distribution of x in Table 5.17,
the mean, variance, and standard deviation are, respectively,
fi = A = .9 car
o- 2 = A = .9
o- = VA = VS = .949 car
EXERCISES
CONCEPTS AND PROCEDURES
5.79 What are the conditions that must be satisfied to apply the Poisson probability distribution?
5.80 What is the parameter of the Poisson probability distribution, and what does it mean?
5.81 Using the Poisson formula, find the following probabilities,
a. P(x < 1) for A = 5 b. P(x = 2) for A = 2.5
Verify these probabilities using Table III of Appendix C.
5.82 Using the Poisson formula, find the following probabilities,
a. P(x < 2) for A = 3 b. P(x = 8) for A = 5.5
Verify these probabilities using Table III of Appendix C.
5.83 Let x be a Poisson random variable. Using the Poisson probabilities table, write the probability dis-
tribution of x for each of the following. Find the mean, variance, and standard deviation for each of these
probability distributions. Draw a graph for each of these probability distributions.
a. A = 1.3 b. A = 2.1
5.8 The Poisson Probability Distribution 237
5.84 Let x be a Poisson random variable. Using the Poisson probabilities table, write the probability dis-
tribution of x for each of the following. Find the mean, variance, and standard deviation for each of these
probability distributions. Draw a graph for each of these probability distributions,
a. A = .6 b. A = 1.8
■ APPLICATIONS
5.85 A household receives an average of 1 .7 pieces of junk mail per day. Find the probability that this
household will receive exactly 3 pieces of junk mail on a certain day. Use the Poisson probability distri-
bution formula.
5.86 A commuter airline receives an average of 9.7 complaints per day from its passengers. Using
the Poisson formula, find the probability that on a certain day this airline will receive exactly 6
complaints.
5.87 On average, 5.4 shoplifting incidents occur per week at an electronics store. Find the probability that
exactly 3 such incidents will occur during a given week at this store.
5.88 On average, 12.5 rooms stay vacant per day at a large hotel in a city. Find the probability that on a
given day exactly 3 rooms will be vacant. Use the Poisson formula.
5.89 A university police department receives an average of 3.7 reports per week of lost student
ID cards.
a. Find the probability that at most 1 such report will be received during a given week by this po-
lice department. Use the Poisson probability distribution formula.
b. Using the Poisson probabilities table, find the probability that during a given week the number of
such reports received by this police department is
i. 1 to 4 ii. at least 6 iii. at most 3
5.90 A large proportion of small businesses in the United States fail during the first few years of opera-
tion. On average, 1.6 businesses file for bankruptcy per day in a particular large city.
a. Using the Poisson formula, find the probability that exactly 3 businesses will file for bankruptcy
on a given day in this city.
b. Using the Poisson probabilities table, find the probability that the number of businesses that will
file for bankruptcy on a given day in this city is
i. 2 to 3 ii. more than 3 iii. less than 3
5.91 Despite all efforts by the quality control department, the fabric made at Benton Corporation always
contains a few defects. A certain type of fabric made at this corporation contains an average of .5 defects
per 500 yards.
a. Using the Poisson formula, find the probability that a given piece of 500 yards of this fabric will
contain exactly 1 defect.
b. Using the Poisson probabilities table, find the probability that the number of defects in a given
500-yard piece of this fabric will be
i. 2 to 4 ii. more than 3 iii. less than 2
5.92 The number of students who login to a randomly selected computer in a college computer lab
follows a Poisson probability distribution with a mean of 19 students per day.
a. Using the Poisson probability distribution formula, determine the probability that exactly 12 stu-
dents will login to a randomly selected computer at this lab on a given day.
b. Using the Poisson probability distribution table, determine the probability that the number of stu-
dents who will login to a randomly selected computer at this lab on a given day is
i. from 13 to 16 ii. fewer than 8
5.93 An average of 4.8 customers come to Columbia Savings and Loan every half hour.
a. Find the probability that exactly 2 customers will come to this savings and loan during a given
hour.
b. Find the probability that during a given hour, the number of customers who will come to this
savings and loan is
i. 2 or fewer ii. 10 or more
5.94 Although Borok's Electronics Company has no openings, it still receives an average of 3.2 unso-
licited applications per week from people seeking jobs.
a. Using the Poisson formula, find the probability that this company will receive no applications
next week.
238 Chapter 5 Discrete Random Variables and Their Probability Distributions
b. Let x denote the number of applications this company will receive during a given week.
Using the Poisson probabilities table from Appendix C, write the probability distribution table
of x.
c. Find the mean, variance, and standard deviation of the probability distribution developed in part b.
5.95 An insurance salesperson sells an average of 1.4 policies per day.
a. Using the Poisson formula, find the probability that this salesperson will sell no insurance policy
on a certain day.
b. Let x denote the number of insurance policies that this salesperson will sell on a given day. Us-
ing the Poisson probabilities table, write the probability distribution of x.
c. Find the mean, variance, and standard deviation of the probability distribution developed in
part b.
5.96 An average of .8 accidents occur per day in a particular large city.
a. Find the probability that no accident will occur in this city on a given day.
b. Let x denote the number of accidents that will occur in this city on a given day. Write the prob-
ability distribution of x.
c. Find the mean, variance, and standard deviation of the probability distribution developed in
part b.
*5.97 On average, 20 households in 50 own answering machines.
a. Using the Poisson formula, find the probability that in a random sample of 50 households, ex-
actly 25 will own answering machines.
b. Using the Poisson probabilities table, find the probability that the number of households in 50
who own answering machines is
i. at most 12 ii. 13 to 17 iii. at least 30
*5.98 Twenty percent of the cars passing through a school zone are exceeding the speed limit by more
than 10 mph.
a. Using the Poisson formula, find the probability that in a random sample of 100 cars passing through
this school zone, exactly 25 will exceed the speed limit by more than 10 mph.
b. Using the Poisson probabilities table, find the probability that the number of cars exceeding the
speed limit by more than 10 mph in a random sample of 100 cars passing through this school
zone is
i. at most 8 ii. 15 to 20 iii. at least 30
USES AND MISUSES...
1. PUT ON YOUR GAME FACE
Gambling would be nothing without probability. A gambler always
has a positive probability of winning. Unfortunately, the house always
plays with better odds. A classic discrete probability distribution ap-
plies to the hands in straight poker. Using the tools you have learned
in this chapter and a bit of creativity, you can derive the probability
of being dealt a certain hand. However, this probability distribution
is only going to be of limited use when you begin to play poker.
The hands in descending order of rank and increasing order of
probability are straight flush, four-of-a-kind, full house, flush, straight,
three-of-a-kind, two pair, pair, and high cards. To begin, let us deter-
mine how many hands there are. As we know, there are 52 cards in
a deck, and any 5 cards can be a valid hand. Using the combinations
notation, we see that there are 52 C 5 or 2,598,960 hands.
We can count the highest hands based on their composition.
The straight flush is any 5 cards in rank order from the same suit. Be-
cause an ace can be high or low, there are 10 straight flushes per
suit. Because there are 4 suits, this gives us 40 straight flushes. Once
you have chosen your rank for four of a kind, for example, a jack,
there are 52 - 4 = 48 remaining cards. Hence, there are 13 x 48 =
624 possible four-of-a-kind hands.
The rest of the hands require us to use the combinations nota-
tion to determine their numbers. A full house is three of a kind and
a pair (for example, three kings and a pair of 7s). There are 13 choices
for three of a kind (for example, three aces, three kings, and so on);
then there are 4 C 3 = 4 ways to choose each set of three of a kind
from 4 cards (for example, 3 kings out of 4). Once three cards of a
kind have been selected, there are 12 possibilities for a pair, and 4 C 2 =
6 ways to choose any 2 cards for a pair out of 4 cards (e.g., two 9s
out of 4). Thus, there are 13 x 4 x 12 x 6 = 3744 full houses.
A flush is five cards drawn from the same suit. Hence, we have 4
suits multiplied by 13 C 5 ways to choose the members, which gives
5148 flushes. However, 40 of those are straight flushes, so 5108
flushes are not straight flushes. For brevity, we omit the calculation
Uses and Misuses 239
of the remainder of the hands and present the results and the prob-
ability of being dealt the hand in the table below.
Number of Hands
Probability
Straight flush
40
.0000154
Four of a kind
624
.0002401
Full house
3744
.0014406
Flush
5108
.0019654
Straight
10,200
.0039246
Three of a kind
54,912
.0211285
Two pair
123,552
.0475390
Pair
1,098,240
.4225690
High card
1,302,540
.5011774
Total
2,598,960
1 .000000
Memorizing this table is only the beginning of poker. Any table
entry represents the probability that the five cards you have been dealt
constitute one of the nine poker hands. Suppose that you are playing
poker with four people and you are dealt a pair of sevens. The prob-
ability that you were dealt that hand is .4225690, but that is not the
probability in which you are interested. You want to know the proba-
bility that the pair of sevens that you hold will beat the hands your op-
ponents were dealt. Despite your intimate knowledge of the probabil-
ity of your hand, the above table gives information for only one player.
Be very careful when working with probability distributions, and make
sure you understand exactly what the probabilities represent.
2. ODDS AND PROBABILITIES
In the Uses and Misuses section of Chapter 4, we discussed the fact
that casinos and odds makers do not necessarily follow the basic
probability rules when determining odds for sports betting. However,
the goal for casinos is to make money, on average, from all forms of
betting, including sports gambling. As we are about to show, failure
to meet the basic probability rules does not negatively impact the
profit one expects to make.
When a person places a Si bet on a team, there are two pos-
sibilities: (1) the casino or betting agency keeps the dollar if the bet-
tor loses; (2) the casino or betting agency pays the bettor the amount
based on the odds. For example, the accompanying table (reproduced
from the Uses and Misuses section of Chapter 4) shows that the odds
for the Detroit Tigers to win the 2009 World Series are 1 :30. Suppose
a bettor places a SI bet on the Detroit Tigers. The bettor will win a
net amount of S29 (= S30 - Si) if Detroit wins the World Series,
and the bettor will lose Si if Detroit does not win the World Series.
Similarly, a person placing a Si bet on the Chicago Cubs will either
win a net amount of S(15/2 - 1) = S6.50 or lose the SI that he or
she bet on the Chicago Cubs.
Note that the table of odds that we reproduce here has a new
column that lists the expected (average) payout for a Si bet on each
team. The expected payout for each team is calculated by multiply-
ing the probabilities of the two outcomes (winning and not winning)
by the respective amounts won/lost and adding these products. For
example, for the Minnesota Twins, the probability of winning the
World Series is 1/17, and the probability of not winning is 16/17,
based on the odds of 1:16. For a Si bet, the bettor will win a net
amount of SI 5 (Si 6 minus the Si bet) if the Twins win, and the bet-
tor will lose Si if the Twins do not win. Thus, the expected net win-
nings of the bettor on a Si bet on the Twins is Sl5(l/17) +
S(-l)(16/17) = -SO/17) = -S.0588. In other words, the bettor is
expected to lose, on average, S.0588 on a Si bet on the Twins. There-
fore, the bettor is expected to receive a payout of SI - S.0588 =
S.9412 on a Si bet. This means that all people who bet on the Twins,
on average, are paid back S.9412 per dollar bet and they lose, on av-
erage, S.0588 per dollar bet, which is the casino's profit.
Team
Current Odds
Expected Payout
Team
Current Odds
Expected Payout
Arizona Diamondbacks
1:12
S.9231
Milwaukee Brewers
1:28
S.9655
Atlanta Braves
1:60
S.9836
Minnesota Twins
1:16
S.9412
Baltimore Orioles
1:80
S.9877
New York Mets
1:16
S.9412
Boston Red Sox
2:13
S.8667
New York Yankees
1:9
S.9000
Chicago Cubs
2:15
S.8824
Oakland Athletics
1:40
S.9756
Chicago White Sox
1:30
S.9677
Philadelphia Phillies
1:12
S.9231
Cincinnati Reds
1:80
S.9877
Pittsburgh Pirates
1:150
S.9934
Cleveland Indians
1:25
S.9615
San Diego Padres
1:100
S.9901
Colorado Rockies
1:38
S.9744
San Francisco Giants
1:59
S.9833
Detroit Tigers
1:30
S.9677
Seattle Mariners
1:80
S.9877
Florida Marlins
1:20
S0.9524
St. Louis Cardinals
1:18
S0.9474
Houston Astros
1:40
S0.9756
Tampa Bay Devil Rays
1:10
S0.9091
Kansas City Royals
1:125
S0.9921
Texas Rangers
1:75
S0.9868
Los Angeles Angels
1:8
S0.8889
Toronto Blue Jays
1:35
S0.9722
Los Angeles Dodgers
1:8
S0.8889
Washington Nationals
1:125
S0.9921
240 Chapter 5 Discrete Random Variables and Their Probability Distributions
If we add the expected payouts for all teams, we obtain the ex-
pected payout on a Si bet on all 30 teams. The sum of the expected
payouts listed in the table is $28.6091. However, the casino received
$30 in $1 bets on each of the 30 teams. Thus, the casino's expected
profit is $30 - $28.6091 = $1.3909. Remember that expected values
are based on a large number of repetitions (bets, in this case). If the
casino received only one $1 bet on each team, then the casino would
be nervous because if any of 14 teams with odds of 1:31 or worse
were to win, the casino would lose money. Even if one million peo-
ple place $1 bets, the casino can still lose money if 6667 or more peo-
ple bet on the Pittsburgh Pirates and this team wins the 2009 World
Series. However, it is highly unlikely that the casino will lose money.
Having said that, the casinos like to offer odds that will encour-
age people to place bets on the teams that are least likely to win, such
as the 2009 Pittsburgh Pirates or Washington Nationals. Casinos know
that if people bet on these teams, it is very unlikely that casinos will
have to pay these bettors. Many people like to place small bets that
have the (extremely small) possibility of a big payoff, which is often
the allure of the multistate lotteries, such as the Powerball and Mega
Millions games. Remember that the odds given in the above table are
at a snapshot in time. These odds are adjusted as the season goes on.
If a team with odds of 1 :150 gets worse, the odds will also get worse
(e.g., to 1 :500 from 1 :150). Similarly, teams that are in first place and
continue to play very well will have their odds improve.
Glossary
Bernoulli trial One repetition of a binomial experiment. Also
called a trial.
Binomial experiment An experiment that contains n identical tri-
als such that each of these n trials has only two possible outcomes,
the probabilities of these two outcomes remain constant for each
trial, and the trials are independent.
Binomial parameters The total trials n and the probability of suc-
cess p for the binomial probability distribution.
Binomial probability distribution The probability distribution
that gives the probability of x successes in n trials when the proba-
bility of success is p for each trial of a binomial experiment.
Combinations The number of ways x elements can be selected
from n elements. Here order of selection is not important.
Continuous random variable A random variable that can assume
any value in one or more intervals.
Discrete random variable A random variable whose values are
countable.
Factorial Denoted by the symbol !. The product of all the integers
from a given number to 1. For example, n\ (read as "/? factorial")
represents the product of all the integers from n to 1 .
Hypergeometric probability distribution The probability distri-
bution that is applied to determine the probability of x successes in
n trials when the trials are not independent.
Mean of a discrete random variable The mean of a discrete ran-
dom variable x is the value that is expected to occur per repetition,
on average, if an experiment is performed a large number of times.
The mean of a discrete random variable is also called its expected
value.
Permutations Number of arrangements of x items selected from
n items. Here order of selection is important.
Poisson parameter The average occurrences, denoted by A, dur-
ing an interval for a Poisson probability distribution.
Poisson probability distribution The probability distribution that
gives the probability of x occurrences in an interval when the aver-
age occurrences in that interval are A.
Probability distribution of a discrete random variable A list of
all the possible values that a discrete random variable can assume
and their corresponding probabilities.
Random variable A variable, denoted by x, whose value is deter-
mined by the outcome of a random experiment. Also called a chance
variable.
Standard deviation of a discrete random variable A measure of
spread for the probability distribution of a discrete random variable.
Supplementary Exercises
5.99 Let x be the number of cars that a randomly selected auto mechanic repairs on a given day. The fol-
lowing table lists the probability distribution of x.
X
2
3
4
5
6
P(x)
.05
.22
.40
.23
.10
Find the mean and standard deviation of x. Give a brief interpretation of the value of the mean.
Supplementary Exercises 241
5.100 Let x be the number of emergency root canal surgeries performed by Dr. Sharp on a given Monday.
The following table lists the probability distribution of x.
X
1
2
3
4
5
P(x)
.13
.28
.30
.17
.08
.04
Calculate the mean and standard deviation of x. Give a brief interpretation of the value of the mean.
5.101 Based on its analysis of the future demand for its products, the financial department at Tipper Cor-
poration has determined that there is a .17 probability that the company will lose $1.2 million during the
next year, a .21 probability that it will lose $.7 million, a .37 probability that it will make a profit of
$.9 million, and a .25 probability that it will make a profit of $2.3 million.
a. Let x be a random variable that denotes the profit earned by this corporation during the next
year. Write the probability distribution of x.
b. Find the mean and standard deviation of the probability distribution of part a. Give a brief in-
terpretation of the value of the mean.
5.102 GESCO Insurance Company charges a $350 premium per annum for a $100,000 life insurance pol-
icy for a 40-year-old female. The probability that a 40-year-old female will die within 1 year is .002.
a. Let x be a random variable that denotes the gain of the company for next year from a
$100,000 life insurance policy sold to a 40-year-old female. Write the probability distribution
of x,
b. Find the mean and standard deviation of the probability distribution of part a. Give a brief in-
terpretation of the value of the mean.
5.103 Spoke Weaving Corporation has eight weaving machines of the same kind and of the same age.
The probability is .04 that any weaving machine will break down at any time. Find the probability that at
any given time
a. all eight weaving machines will be broken down
b. exactly two weaving machines will be broken down
c. none of the weaving machines will be broken down
5.104 At the Bank of California, past data show that 8% of all credit card holders default at some time
in their lives. On one recent day, this bank issued 12 credit cards to new customers. Find the probability
that of these 12 customers, eventually
a. exactly 3 will default b. exactly 1 will default c. none will default
5.105 Maine Corporation buys motors for electric fans from another company that guarantees
that at most 5% of its motors are defective and that it will replace all defective motors at no cost to
Maine Corporation. The motors are received in large shipments. The quality control department at
Maine Corporation randomly selects 20 motors from each shipment and inspects them for being good
or defective. If this sample contains more than 2 defective motors, the entire shipment is rejected.
a. Using the appropriate probabilities table from Appendix C, find the probability that a given
shipment of motors received by Maine Corporation will be accepted. Assume that 5% of all
motors received are defective.
b. Using the appropriate probabilities table from Appendix C, find the probability that a given
shipment of motors received by Maine Corporation will be rejected.
5.106 One of the toys made by Dillon Corporation is called Speaking Joe, which is sold only by mail.
Consumer satisfaction is one of the top priorities of the company's management. The company guar-
antees a refund or a replacement for any Speaking Joe toy if the chip that is installed inside becomes
defective within 1 year from the date of purchase. It is known from past data that 10% of these chips
become defective within a 1-year period. The company sold 15 Speaking Joes on a given day.
a. Let x denote the number of Speaking Joes in these 15 that will be returned for a refund or a re-
placement within a 1-year period. Using the appropriate probabilities table from Appendix C,
obtain the probability distribution of x and draw a graph of the probability distribution. Deter-
mine the mean and standard deviation of x.
b. Using the probability distribution constructed in part a, find the probability that exactly 5
of the 15 Speaking Joes will be returned for a refund or a replacement within a 1-year
period.
5.107 In a list of 15 households, 9 own homes and 6 do not own homes. Four households are randomly
selected from these 15 households. Find the probability that the number of households in these 4 who own
homes is
a. exactly 3 b. at most 1 c. exactly 4
242 Chapter 5 Discrete Random Variables and Their Probability Distributions
5.108 Twenty corporations were asked whether or not they provide retirement benefits to their employ-
ees. Fourteen of the corporations said they do provide retirement benefits to their employees, and 6 said
they do not. Five corporations are randomly selected from these 20. Find the probability that
a. exactly 2 of them provide retirement benefits to their employees.
b. none of them provides retirement benefits to their employees.
c. at most one of them provides retirement benefits to employees.
5.109 Uniroyal Electronics Company buys certain parts for its refrigerators from Bob's Corporation. The
parts are received in shipments of 400 boxes, each box containing 16 parts. The quality control depart-
ment at Uniroyal Electronics first randomly selects 1 box from each shipment and then randomly selects
4 parts from that box. The shipment is accepted if at most 1 of the 4 parts is defective. The quality con-
trol inspector at Uniroyal Electronics selected a box from a recently received shipment of such parts.
Unknown to the inspector, this box contains 3 defective parts.
a. What is the probability that this shipment will be accepted?
b. What is the probability that this shipment will not be accepted?
5.110 Alison Bender works for an accounting firm. To make sure her work does not contain errors, her
manager randomly checks on her work. Alison recently filled out 12 income tax returns for the company's
clients. Unknown to anyone, 2 of these 12 returns have minor errors. Alison's manager randomly selects
3 returns from these 12 returns. Find the probability that
a. exactly 1 of them contains errors.
b. none of them contains errors.
c. exactly 2 of them contain errors.
5.111 The student health center at a university treats an average of seven cases of mononucleosis per day
during the week of final examinations.
a. Using the appropriate formula, find the probability that on a given day during the finals week
exactly four cases of mononucleosis will be treated at this health center.
b. Using the appropriate probabilities table from Appendix C, find the probability that on a given
day during the finals week the number of cases of mononucleosis treated at this health center
will be
i. at least 7 ii. at most 3 iii. 2 to 5
5.112 An average of 6.3 robberies occur per day in a large city.
a. Using the Poisson formula, find the probability that on a given day exactly 3 robberies will oc-
cur in this city.
b. Using the appropriate probabilities table from Appendix C, find the probability that on a given
day the number of robberies that will occur in this city is
i. at least 12 ii. at most 3 iii. 2 to 6
5.113 An average of 1.4 private airplanes arrive per hour at an airport.
a. Find the probability that during a given hour no private airplane will arrive at this airport.
b. Let x denote the number of private airplanes that will arrive at this airport during a given hour.
Write the probability distribution of x.
5.114 A high school boys' basketball team averages 1.2 technical fouls per game.
a. Using the appropriate formula, find the probability that in a given basketball game this team
will commit exactly 3 technical fouls.
b. Let x denote the number of technical fouls that this team will commit during a given basketball
game. Using the appropriate probabilities table from Appendix C, write the probability distribu-
tion of x.
Advanced Exercises
5.115 Scott offers you the following game: You will roll two fair dice. If the sum of the two numbers ob-
tained is 2, 3, 4, 9, 10, 11, or 12, Scott will pay you $20. However, if the sum of the two numbers is 5,
6, 7, or 8, you will pay Scott $20. Scott points out that you have seven winning numbers and only four
losing numbers. Is this game fair to you? Should you accept this offer? Support your conclusion with ap-
propriate calculations.
5.116 Suppose the owner of a salvage company is considering raising a sunken ship. If successful, the
venture will yield a net profit of $10 million. Otherwise, the owner will lose $4 million. Let p denote the
probability of success for this venture. Assume the owner is willing to take the risk to go ahead with this
project provided the expected net profit is at least $500,000.
a. If p — .40, find the expected net profit. Will the owner be willing to take the risk with this
probability of success?
Supplementary Exercises 243
b. What is the smallest value of p for which the owner will take the risk to undertake this
project?
5.117 Two teams, A and B, will play a best-of-seven series, which will end as soon as one of the teams
wins four games. Thus, the series may end in four, five, six, or seven games. Assume that each team has
an equal chance of winning each game and that all games are independent of one another. Find the fol-
lowing probabilities.
a. Team A wins the series in four games.
b. Team A wins the series in five games.
c. Seven games are required for a team to win the series.
5.118 York Steel Corporation produces a special bearing that must meet rigid specifications. When the
production process is running properly, 10% of the bearings fail to meet the required specifications. Some-
times problems develop with the production process that cause the rejection rate to exceed 10%. To guard
against this higher rejection rate, samples of 15 bearings are taken periodically and carefully inspected. If
more than 2 bearings in a sample of 15 fail to meet the required specifications, production is suspended
for necessary adjustments.
a. If the true rate of rejection is 10% (that is, the production process is working properly), what is
the probability that the production will be suspended based on a sample of 15 bearings?
b. What assumptions did you make in part a?
5.119 Residents in an inner-city area are concerned about drug dealers entering their neighborhood.
Over the past 14 nights, they have taken turns watching the street from a darkened apartment. Drug
deals seem to take place randomly at various times and locations on the street and average about three
per night. The residents of this street contacted the local police, who informed them that they do not
have sufficient resources to set up surveillance. The police suggested videotaping the activity on the
street, and if the residents are able to capture five or more drug deals on tape, the police will take ac-
tion. Unfortunately, none of the residents on this street owns a video camera and, hence, they would
have to rent the equipment. Inquiries at the local dealers indicated that the best available rate for rent-
ing a video camera is $75 for the first night and $40 for each additional night. To obtain this rate, the
residents must sign up in advance for a specified number of nights. The residents hold a neighborhood
meeting and invite you to help them decide on the length of the rental period. Because it is difficult for
them to pay the rental fees, they want to know the probability of taping at least five drug deals on a
given number of nights of videotaping.
a. Which of the probability distributions you have studied might be helpful here?
b. What assumption(s) would you have to make?
c. If the residents tape for two nights, what is the probability they will film at least five drug
deals?
d. For how many nights must the camera be rented so that there is at least .90 probability that five
or more drug deals will be taped?
5.120 A high school history teacher gives a 50-question multiple-choice examination in which each
question has four choices. The scoring includes a penalty for guessing. Each correct answer is worth
1 point, and each wrong answer costs 1/2 point. For example, if a student answers 35 questions cor-
rectly, 8 questions incorrectly, and does not answer 7 questions, the total score for this student will be
35 - (l/2)(8) = 31.
a. What is the expected score of a student who answers 38 questions correctly and guesses on the
other 12 questions? Assume that the student randomly chooses one of the four answers for
each of the 12 guessed questions.
b. Does a student increase his expected score by guessing on a question if he has no idea what
the correct answer is? Explain.
c. Does a student increase her expected score by guessing on a question for which she can elimi-
nate one of the wrong answers? Explain.
5.121 A baker who makes fresh cheesecakes daily sells an average of five such cakes per day. How many
cheesecakes should he make each day so that the probability of running out and losing one or more sales
is less than .10? Assume that the number of cheesecakes sold each day follows a Poisson probability dis-
tribution. You may use the Poisson probabilities table from Appendix C.
5.122 Suppose that a certain casino has the "money wheel" game. The money wheel is divided into
50 sections, and the wheel has an equal probability of stopping on each of the 50 sections when it is spun.
Twenty- two of the sections on this wheel show a $1 bill, 14 show a $2 bill, 7 show a $5 bill, 3 show a
$10 bill, 2 show a $20 bill, 1 shows a flag, and 1 shows a joker. A gambler may place a bet on any of the
seven possible outcomes. If the wheel stops on the outcome that the gambler bet on, he or she wins. The
net payoffs for these outcomes for $1 bets are as follows.
244 Chapter 5 Discrete Random Variables and Their Probability Distributions
Symbol bet on
$1
$2
$5
$10
$20
Flag
Joker
Payoff (dollars)
1
2
5
10
20
40
40
a. If the gambler bets on the $1 outcome, what is the expected net payoff?
b. Calculate the expected net payoffs for each of the other six outcomes.
c. Which bet(s) is (are) best in terms of expected net payoff? Which is (are) worst?
5.123 A history teacher has given her class a list of seven essay questions to study before the next test.
The teacher announced that she will choose four of the seven questions to give on the test, and each stu-
dent will have to answer three of those four questions.
a. In how many ways can the teacher choose four questions from the set of seven?
b. Suppose that a student has enough time to study only five questions. In how many ways can
the teacher choose four questions from the set of seven so that the four selected questions in-
clude both questions that the student did not study?
c. What is the probability that the student in part b will have to answer a question that he or she
did not study? That is, what is the probability that the four questions on the test will include
both questions that the student did not study?
5.124 Consider the following three games. Which one would you be most likely to play? Which one would
you be least likely to play? Explain your answer mathematically.
Game I: You toss a fair coin once. If a head appears you receive $3, but if a tail appears you have
to pay $1.
Game II: You buy a single ticket for a raffle that has a total of 500 tickets. Two tickets are chosen
without replacement from the 500. The holder of the first ticket selected receives $300,
and the holder of the second ticket selected receives $150.
Game III: You toss a fair coin once. If a head appears you receive $1,000,002, but if a tail appears
you have to pay $1,000,000.
5.125 Brad Henry is a stone products salesman. Let x be the number of contacts he visits on a particular
day. The following table gives the probability distribution of x.
X
1
2
3
4
P(x)
.12
.25
.56
.07
Let y be the total number of contacts Brad visits on two randomly selected days. Write the probability dis-
tribution for y.
5.126 The number of calls that come into a small mail-order company follows a Poisson distribution. Cur-
rently, these calls are serviced by a single operator. The manager knows from past experience that an ad-
ditional operator will be needed if the rate of calls exceeds 20 per hour. The manager observes that 9 calls
came into the mail-order company during a randomly selected 15-minute period.
a. If the rate of calls is actually 20 per hour, what is the probability that 9 or more calls will come
in during a given 15-minute period?
b. If the rate of calls is really 30 per hour, what is the probability that 9 or more calls will come
in during a given 15-minute period?
c. Based on the calculations in parts a and b, do you think that the rate of incoming calls is more
likely to be 20 or 30 per hour?
d. Would you advise the manager to hire a second operator? Explain.
5.127 Many of you probably played the game "Rock, Paper, Scissors" as a child. Consider the following
variation of that game. Instead of two players, suppose three players play this game, and let us call these
players A, B, and C. Each player selects one of these three items — Rock, Paper, or Scissors — independent
of each other. Player A will win the game if all three players select the same item, for example, rock.
Player B will win the game if exactly two of the three players select the same item and the third player
selects a different item. Player C will win the game if every player selects a different item. If Player B
wins the game, he or she will be paid $1. If Player C wins the game, he or she will be paid $3. Assum-
ing that the expected winnings should be the same for each player to make this a fair game, how much
should Player A be paid if he or she wins the game?
5.128 Customers arrive at the checkout counter of a supermarket at an average rate of 10 per hour, and
these arrivals follow a Poisson distribution. Using each of the following two methods, find the probabil-
ity that exactly 4 customers will arrive at this checkout counter during a 2-hour period.
Self-Review Test 245
a. Use the arrivals in each of the two nonoverlapping 1-hour periods and then add these. (Note that
the numbers of arrivals in two nonoverlapping periods are independent of each other.)
b. Use the arrivals in a single 2-hour period.
5.129 Consider the Uses and Misuses section in this chapter on poker, where we learned how to calcu-
late the probabilities of specific poker hands. Find the probability of being dealt
a. three of a kind b. two pairs c. one pair
Self-Review Test
1. Briefly explain the meaning of a random variable, a discrete random variable, and a continuous ran-
dom variable. Give one example each of a discrete and a continuous random variable.
2. What name is given to a table that lists all the values that a discrete random variable x can assume
and their corresponding probabilities?
3. For the probability distribution of a discrete random variable, the probability of any single value of
x is always
a. in the range to 1 b. 1.0 c. less than zero
4. For the probability distribution of a discrete random variable, the sum of the probabilities of all pos-
sible values of x is always
a. greater than 1 b. 1.0 c. less than 1.0
5. The number of combinations of 10 items selected 7 at a time is
a. 120 b. 200 c. 80
6. State the four conditions of a binomial experiment. Give one example of such an experiment.
7. The parameters of the binomial probability distribution are
a. n, p, and q b. n and p c. n, p, and x
8. The mean and standard deviation of a binomial probability distribution with = 25 and p = .20 are
a. 5 and 2 b. 8 and 4 c. 4 and 3
9. The binomial probability distribution is symmetric if
a. p < .5 b. p = .5 c. p > .5
10. The binomial probability distribution is skewed to the right if
a. p < .5 b. p = .5 c. p > .5
11. The binomial probability distribution is skewed to the left if
a. p < .5 b. p = .5 c. p > .5
12. Briefly explain when a hypergeometric probability distribution is used. Give one example of a
hypergeometric probability distribution.
13. The parameter/parameters of the Poisson probability distribution is/are
a. A b. A and x c. A and e
14. Describe the three conditions that must be satisfied to apply the Poisson probability distribution.
15. Let x be the number of homes sold per week by all four real estate agents who work at a realty of-
fice. The following table lists the probability distribution of x.
X
1
2
3
4
5
P(x)
.15
.24
.29
.14
.10
.08
Calculate the mean and standard deviation of x. Give a brief interpretation of the value of the mean.
16. According to a survey, 60% of adults believe that all college students should be required to perform
a specified number of hours of community service to graduate. Assume that this percentage is true for the
current population of all adults.
a. Find the probability that the number of adults in a random sample of 12 who hold this view is
i. exactly 8 (use the appropriate formula)
ii. at least 6 (use the appropriate table from Appendix C)
iii. less than 4 (use the appropriate table from Appendix C)
246
Chapter 5 Discrete Random Variables and Their Probability Distributions
b. Let x be the number of adults in a random sample of 12 who believe that all college students
should be required to perform a specified number of hours of community service to graduate.
Using the appropriate table from Appendix C, write the probability distribution of x. Find the
mean and standard deviation of x.
17. The Red Cross honors and recognizes its best volunteers from time to time. One of the Red Cross
offices has received 12 nominations for the next group of 4 volunteers to be recognized. Eight of these 12
nominated volunteers are female. If the Red Cross office decides to randomly select 4 names out of these
12 nominated volunteers, find the probability that of these 4 volunteers
a. exactly 3 are female.
b. exactly 1 is female.
c. at most 1 is female.
18. The police department in a large city has installed a traffic camera at a busy intersection. Any car
that runs a red light will be photographed with its license plate visible, and the driver will receive a cita-
tion. Suppose that during the morning rush hour of weekdays, an average of 10 drivers are caught run-
ning the red light per day by this system.
a. Find the probability that during the morning rush hour on a given weekday this system will catch
i. exactly 14 drivers (use the appropriate formula)
ii. at most 7 drivers (use the appropriate table from Appendix C)
iii. 13 to 18 drivers (use the appropriate table from Appendix C)
b. Let x be the number of drivers caught by this system during the morning rush hour on a given
weekday. Write the probability distribution of x. Use the appropriate table from Appendix C.
19. The binomial probability distribution is symmetric when p = .50, it is skewed to the right when p < .50,
and it is skewed to the left when p > .50. Illustrate these three cases by writing three probability distri-
butions and graphing them. Choose any values of n and p and use the table of binomial probabilities (Table
I of Appendix C).
Mini-Projects
■ MINI-PROJECT 5-1
Consider the NBA data given in Data Set III that accompanies this text.
a. What proportion of these players are less than 74 inches tall?
b. Suppose a random sample of 25 of these players is taken, and x is the number of players in the
sample who are less than 74 inches tall. Find P(x = 0), P(x = 1), P(x = 2), P(x = 3), P(x = 4),
and P(x = 5).
c. Note that x in part b has a binomial distribution with jj, = np. Use the Poisson probabilities table of
Appendix C to approximate P(x = 0), P(x = 1), P(x = 2), P(x = 3), P(x = 4), and P{x = 5).
d. Are the probabilities of parts b and c consistent, or is the Poisson approximation inaccurate? Ex-
plain why.
■ MINI-PROJECT 5-2
Obtain information on the odds and payoffs of one of the instant lottery games in your state or a nearby
state. Let the random variable x be the net amount won on one ticket (payoffs minus purchase price).
Using the concepts presented in this chapter, find the probability distribution of x. Then calculate the
mean and standard deviation of x. What is the player's average net gain (or loss) per ticket purchased?
■ MINI-PROJECT 5-3
For this project, first collect data by doing the following. Select an intersection in your town that is con-
trolled by traffic light. For a specific time period (e.g., 9-10 a.m. or 5-6 p.m.), count the number of cars
that arrive at that intersection from any one direction during each light cycle. Make sure that you do not
count a car twice if it has to sit through two red lights before getting through the intersection. Perform the
following tasks with your data.
a. Create a graphical display of your data. Describe the shape of the distribution. Also discuss which
of the following graphs is more useful for displaying the data you collected: a dotplot, a bar graph,
or a histogram.
Decide for Yourself 247
b. Calculate the mean and variance for your data for light cycles. Note that your sample size is the
number of light cycles you observed. Do you notice a relationship between these two summary
measures? If so, explain what it is.
c. For each unique number of arrivals in your data, calculate the proportion of light cycles that had
that number of arrivals. For example, suppose you collected these data for 100 light cycles, and
you observed 8 cars arriving for each of 12 light cycles. Then, 12/100 = .12 of the light cycles
had 8 arrivals. Also calculate the theoretical probabilities for each number of arrivals using the
Poisson distribution with A equal to the sample mean that you obtained in part b. How do the two
sets of probabilities compare? Is the Poisson a satisfactory model for your data?
DECIDE FOR YOURSELF
Deciding About Investing
If you are a traditional college student, it is quite likely that your
financial portfolio includes a checking account and, possibly, a sav-
ings account. However, before you know it, you will graduate from
college and take a job. On your first day of work, you will have a
meeting with your personnel/human resource manager to discuss,
among other things, your retirement plans. You may decide to invest
a portion of your earnings in a variety of accounts (usually mutual
funds) with the hope that you will have enough money to carry you
through your golden years. But wait — How does one decide which
mutual funds to invest in? Moreover, how does this relate to the con-
cepts of expected value and variance?
The following table lists the top 10 (as of May 30, 2009) mid-
cap growth mutual funds based on the 5-year average return (Source:
http://biz.yahoo.eom/p/tops/mg.html). The table also lists the stan-
dard deviations of the annual returns for these funds.
By looking at and analyzing the annual returns and the standard
deviations of the annual returns for the mutual funds listed in the
table, some questions arise that you should try to answer.
1. If you decide to invest in a mutual fund based solely on these
average annual returns, which fund would you invest in and why? Is
this a wise decision?
2. The American Century Heritage Inst fund has the highest aver-
age annual return over the 5-year period as shown in the table. Does
this imply that the fund is still doing better than all of the other funds
listed in the table? Why or why not? Do you think this fund will con-
tinue to do better than other funds in the future?
3. By considering both the average annual return and the standard
deviation of the annual returns, why might a person choose to invest
in the BB&T Special Opportunities Equity Inst fund over the
American Century Heritage Inst fund, even though the average annual
return is more than 14% lower for the BB&T Special Opportunities
Equity Inst fund?
4. Which of these funds would you invest in and why?
5. People who are in their 20s and 30s can afford to take more
risks with their investment portfolios because they have plenty of
time to offset short-term losses. However, people who are closer to
retirement age are less likely to take such high risks. Assuming that
the future behavior of the mutual funds is comparable to that of the
past 5 years, which of the mutual funds listed in the table would be
better to invest in if you are in your 20s or 30s and why? What if you
are close to retiring?
Fund Name
Symbol
Annual Return
Standard Deviation
American Century Heritage Inst
ATHIX
6.13%
20.90%
BlackRock U.S. Opportunities Instl
BMCIX
5.99%
18.58%
American Century Heritage Inv
TWHIX
5.91%
20.87%
American Century Heritage A
ATHAX
5.65%
20.89%
BlackRock U.S. Opportunities Svc
BMCSX
5.60%
18.60%
BlackRock U.S. Opportunities Inv A
BMEAX
5.51%
18.57%
BB&T Special Opportunities Equity Inst
BOPIX
5.26%
16.52%
BB&T Special Opportunities Equity A
BOPAX
5.02%
16.52%
American Century Giftrust Inv
TWGTX
4.86%
20.64%
American Century Heritage C
AHGCX
4.86%
20.88%
248 Chapter 5 Discrete Random Variables and Their Probability Distributions
ECHNOLOGY
INSTRUCTION
Combinations, Binomial Distribution and Poisson Distribution
ti DRAW
1. To find the number of ways of choosing x objects out of n, type n, select MATH >PRB
>nCr, then type x and press ENTER.
2. To find the binomial probability of x successes out of n trials, with each trial having a proba-
bility p of success, select DISTR >binompdf(«, p, x), and press ENTER. (See Screen 5.1.)
3. To find the Poisson probability of x successes when mean is A, select DISTR >poissonpdf
screen 5.1 (A, x), and press ENTER.
0: binonpdf (
fl: binomcdf (
Binomial Distribution
(• Probability
C Cumulative probability
r Inverse cumulative probability
Select
Number of trials:
Probability of success:
C Input column:
Optional storage:
(• Input constant:
Optional storage:
fTF
IT
1. To find the probability of x successes in n trials
for a binomial random variable with probability
of success p, select Calc >Probability Distri-
butions >Binomial. In the dialog box, make
sure that Probability is selected, then enter the
number of trials n, as well as the probability p
of success. Select Input constant, and enter
the value of x. (See Screen 5.2.)
If you need to create a table of probabilities
for various values of x, first enter them into a
column. Again select Calc >Probability
Distributions > Binomial and enter n and p.
Now select Input column and enter the name
of the column where you entered x. If you wish
to store the resulting probabilities, enter the
name of a column under Optional storage.
2. To find the probability of x for a Poisson ran-
dom variable, select Calc >Probability Distrib-
utions > Poisson. In the dialog box, make sure
Screen 5.2 mat Probability is selected, then enter the value
of mean A. Select Input constant, and enter x.
If you need to create a table of probabilities for various values of x, first enter them into a
column. Again select Calc >Probability Distributions >Poisson and enter A. Now select
Input column and enter the name of the column where you entered x. If you wish to store
the resulting probabilities, enter the name of a column under Optional storage.
Help
OK
Cancel
A B
C
E
1
2
Number of combinations of 3 objacls out of 10:
3
120
Screen 5.3
1. To find the number of combinations of x objects chosen out of n, type
=COMBIN(n, x). (See Screens 5.3 and 5.4.)
2. To find the probability for a binomial random variable to take the value x out
of n trials with probability of success p, type =BINOMDIST(x, n, p, 0).
3. To find the probability for a Poisson random
variable to take the value x with the average
number of occurrences A,
type =POISSON(x, X, 0).
Screen 5.4
A B
c
D I
1
Number of combinations of 3 objects out of
0:
2
3
=COMBIN(103)
A
| Hcr*iBirf:'r»jmrjef, numbet.chosen)
1 !
Technology Assignments 249
TECHNOLOGY ASSIGNMENTS
TA5.1 Forty-five percent of the adult population in a particular large city are women. A court is to ran-
domly select a jury of 12 adults from the population of all adults of this city.
a. Find the probability that none of the 12 jurors is a woman.
b. Find the probability that at most 4 of the 12 jurors are women.
c. Let x denote the number of women in 12 adults selected for this jury. Obtain the probability distri-
bution of x.
d. Using the probability distribution obtained in part c, find the following probabilities.
i. P(x > 6) ii. P(x < 3) iii. P(2 < x < 7)
TA5.2 According to an NPD Group poll, 63% of 18 to 24-year-old women said they shop only at their
favorite stores (Forbes, November 1, 2004). Assume that this percentage is true for the current population
of such women.
a. Find the probability that in a random sample of 20 such women, exactly 14 will say that they shop
only at their favorite stores.
b. Find the probability that in a random sample of 30 such women, exactly 18 will say that they
shop only at their favorite stores.
c. Find the probability that in a random sample of 25 such women, at most 15 will say that they shop
only at their favorite stores.
d. Find the probability that in a random sample of 40 such women, at least 30 will say that they shop
only at their favorite stores.
TA5.3 A mail-order company receives an average of 40 orders per day.
a. Find the probability that it will receive exactly 55 orders on a certain day.
b. Find the probability that it will receive at most 29 orders on a certain day.
c. Let x denote the number of orders received by this company on a given day. Obtain the probability
distribution of x.
d. Using the probability distribution obtained in part c, find the following probabilities.
i. P(x > 45) ii. P(x < 33) iii. P(36 <x< 52)
TA5.4 A commuter airline receives an average of 13 complaints per week from its passengers. Let x de-
note the number of complaints received by this airline during a given week.
a. Find P(x = 0). If your answer is zero, does it mean that this cannot happen? Explain.
b. Find P(x < 10).
c. Obtain the probability distribution of x.
d. Using the probability distribution obtained in part c, find the following probabilities,
i. P(x > 18) ii. P(x < 9) iii. P(\0 < x < 17)
Chapter
6.1 Continuous Probability
Distribution
Case Study 6-1 Distribution
of Time Taken to Run a
Road Race
6.2 The Normal Distribution
6.3 The Standard Normal
Distribution
6.4 Standardizing a Normal
Distribution
6.5 Applications of the
Normal Distribution
6.6 Determining the z and
x Values When an Area
Under the Normal
Distribution Curve Is
Known
6.7 The Normal
Approximation to the
Binomial Distribution
Continuous Random Variables
and the Normal Distribution
Have you ever participated in a road race? If you have, where did you stand in comparison to
the other runners? Do you think the time taken to finish a road race varies as much among run-
ners as the runners themselves? See Case Study 6-1 for the distribution of times for runners who
completed the Manchester (Connecticut) Road Race in 2008.
Discrete random variables and their probability distributions were presented in Chapter 5. Section 5.1
defined a continuous random variable as a variable that can assume any value in one or more intervals.
The possible values that a continuous random variable can assume are infinite and uncountable.
For example, the variable that represents the time taken by a worker to commute from home to work
is a continuous random variable. Suppose 5 minutes is the minimum time and 130 minutes is the
maximum time taken by all workers to commute from home to work. Let x be a continuous random
variable that denotes the time taken to commute from home to work by a randomly selected worker.
Then x can assume any value in the interval 5 to 130 minutes. This interval contains an infinite num-
ber of values that are uncountable.
A continuous random variable can possess one of many probability distributions. In this chapter,
we discuss the normal probability distribution and the normal distribution as an approximation to the
binomial distribution.
250
6.1 Continuous Probability Distribution
6.1 Continuous Probability Distribution
In Chapter 5, we defined a continuous random variable as a random variable whose values
are not countable. A continuous random variable can assume any value over an interval or in-
tervals. Because the number of values contained in any interval is infinite, the possible number
of values that a continuous random variable can assume is also infinite. Moreover, we cannot
count these values. In Chapter 5, it was stated that the life of a battery, heights of people, time
taken to complete an examination, amount of milk in a gallon, weights of babies, and prices of
houses are all examples of continuous random variables. Note that although money can be
counted, all variables involving money usually are considered to be continuous random vari-
ables. This is so because a variable involving money often has a very large number of outcomes.
Suppose 5000 female students are enrolled at a university, and x is the continuous random
variable that represents the heights of these female students. Table 6.1 lists the frequency and
relative frequency distributions of x.
Table 6.1 Frequency and Relative Frequency
Distributions of Heights of Female Students
Height of a Female
Student (inches) Relative
X
/
Frequency
60 to less
than
61
90
.018
61 to less
than
62
170
.034
62 to less
than
63
460
.092
63 to less
than
64
750
.150
64 to less
than
65
970
.194
65 to less
than
66
760
.152
66 to less
than
67
640
.128
67 to less
than
68
440
.088
68 to less
than
69
320
.064
69 to less
than
70
220
.044
70 to less
than
71
180
.036
N = 5000
Sum = 1.0
The relative frequencies given in Table 6.1 can be used as the probabilities of the respec-
tive classes. Note that these are exact probabilities because we are considering the population
of all female students.
Figure 6.1 displays the histogram and polygon for the relative frequency distribution of
Table 6.1. Figure 6.2 shows the smoothed polygon for the data of Table 6.1. The smoothed
Figure 6.1 Histogram and polygon
for Table 6.1.
252 Chapter 6 Continuous Random Variables and the Normal Distribution
Figure 6.2 Probability distribution curve for heights.
60 61 62 63 64 65 66 67 68 69 70 71
polygon is an approximation of the probability distribution curve of the continuous random
variable x. Note that each class in Table 6.1 has a width equal to 1 inch. If the width of classes
is more than 1 unit, we first obtain the relative frequency densities and then graph these rel-
ative frequency densities to obtain the distribution curve. The relative frequency density of a
class is obtained by dividing the relative frequency of that class by the class width. The rel-
ative frequency densities are calculated to make the sum of the areas of all rectangles in the
histogram equal to 1.0. Case Study 6-1, which appears later in this section, illustrates this
procedure. The probability distribution curve of a continuous random variable is also called
its probability density function.
The probability distribution of a continuous random variable possesses the following two
characteristics.
1. The probability that x assumes a value in any interval lies in the range to 1.
2. The total probability of all the (mutually exclusive) intervals within which x can assume a
value is 1.0.
The first characteristic states that the area under the probability distribution curve of a contin-
uous random variable between any two points is between and 1, as shown in Figure 6.3. The
second characteristic indicates that the total area under the probability distribution curve of a
continuous random variable is always 1.0, or 100%, as shown in Figure 6.4.
Figure 6.3 Area under a curve between two points.
x = a x = b x
Figure 6.4 Total area under a probability distribution
curve.
.V
The probability that a continuous random variable x assumes a value within a certain in-
terval is given by the area under the curve between the two limits of the interval, as shown in
6.1 Continuous Probability Distribution 253
Figure 6.5. The shaded area under the curve from a to b in this figure gives the probability that
x falls in the interval a to b\ that is,
P(a < x < b) = Area under the curve from a to b
Note that the interval a < x < states that x is greater than or equal to a but less than or equal
to b.
a b x
Figure 6.5 Area under the curve as probability.
Reconsider the example on the heights of all female students at a university. The prob-
ability that the height of a randomly selected female student from this university lies in the
interval 65 to 68 inches is given by the area under the distribution curve of the heights
of all female students from x = 65 to x = 68, as shown in Figure 6.6. This probability is
written as
P(65 < x < 68)
which states that x is greater than or equal to 65 but less than or equal to 68.
65 68 x
Figure 6.6 Probability that x lies in the interval
65 to 68.
For a continuous probability distribution, the probability is always calculated for an in-
terval. For example, in Figure 6.6, the interval representing the shaded area is from
65 to 68. Consequently, the shaded area in that figure gives the probability for the interval
65 < x < 68.
The probability that a continuous random variable x assumes a single value is always
zero. This is so because the area of a line, which represents a single point, is zero. For exam-
ple, if x is the height of a randomly selected female student from that university, then the prob-
ability that this student is exactly 67 inches tall is zero; that is,
P(x = 67) =
This probability is shown in Figure 6.7. Similarly, the probability for x to assume any other sin-
gle value is zero.
In general, if a and b are two of the values that x can assume, then
P(a) = and P(b) =
254
Chapter 6 Continuous Random Variables and the Normal Distribution
67 x
Figure 6.7 The probability of a single value of
x is zero.
From this we can deduce that for a continuous random variable,
P(a < jc < b) = P(a < x < b)
In other words, the probability that x assumes a value in the interval a to b is the same whether
or not the values a and b are included in the interval. For the example on the heights of fe-
male students, the probability that a randomly selected female student is between 65 and 68
inches tall is the same as the probability that this female is 65 to 68 inches tall. This is shown
in Figure 6.8.
— Shaded area gives
the probability
P(65<*<68)
as well as
\P(65 <x< 68)
65 68 x
Figure 6.8 Probability "from 65 to 68" and "between
65 and 68."
Note that the interval "between 65 and 68" represents "65 < x < 68" and it does not in-
clude 65 and 68. On the other hand, the interval "from 65 to 68" represents "65 < jc < 68"
and it does include 65 and 68. However, as mentioned previously, in the case of a continu-
ous random variable, both of these intervals contain the same probability or area under the
curve.
Case Study 6-1 on the next page describes how we obtain the probability distribution curve
of a continuous random variable.
6.2 The Normal Distribution
The normal distribution is one of the many probability distributions that a continuous random
variable can possess. The normal distribution is the most important and most widely used of all
probability distributions. A large number of phenomena in the real world are normally distrib-
uted either exactly or approximately. The continuous random variables representing heights and
weights of people, scores on an examination, weights of packages (e.g., cereal boxes, boxes of
cookies), amount of milk in a gallon, life of an item (such as a light-bulb or a television set),
and time taken to complete a certain job have all been observed to have a (approximate) nor-
mal distribution.
The following table gives the frequency and relative frequency distributions for the time (in minutes) taken
to complete the Manchester Road Race (held on November 27, 2008) by a total of 10,431 participants who
finished that race. This event is held every year on Thanksgiving Day in Manchester, Connecticut. The total
distance of the course is 4.748 miles. The relative frequencies in the following table are used to construct
the histogram and polygon in Figure 6.9.
Liass
Frequency
Relative
Frequency
20 to less than
25
30
.0029
25 to less than
30
208
.0199
30 to less than
35
771
.0739
35 to less than
40
1099
.1054
40 to less than
45
1137
.1090
45 to less than
50
1660
.1591
50 to less than
55
1751
.1679
55 to less than
60
1346
.1290
60 to less than
65
800
.0767
65 to less than
70
419
.0402
70 to less than
75
313
.0300
75 to less than
80
238
.0228
80 to less than
85
178
.0171
85 to less than
90
178
.0171
90 to less than
95
149
.0143
95 to less than
100
107
.0103
100 to less than
105
23
.0022
105 to less than
110
16
.0015
1 1 to less than
115
8
.0008
If = 10,431
Sum = 1.0001
DISTRIBU-
TION OF
TIME
TAKEN TO
RUN A
ROAD RACE
To derive the probability distribution curve for these data, we calculate the relative frequency densities by
dividing the relative frequencies by the class widths. The width of each class in the above table is 5. By di-
viding the relative frequencies by 5, we obtain the relative frequency densities, which are recorded in the
next table. Using the relative frequency densities, we draw a histogram and smoothed polygon, as shown
in Figure 6.10. The curve in this figure is the probability distribution curve for the Road Race data.
Note that the areas of the rectangles in Figure 6.9 do not give probabilities (which are approximated
by relative frequencies). Rather, it is the heights of these rectangles that give the probabilities. This is so
CONTINUED
because the base of each rectangle is 5 in this histogram. Consequently, the area of any rectangle is given
by its height multiplied by 5. Thus, the total area of all the rectangles in Figure 6.9 is 5.0, not 1.0. How-
ever, in Figure 6.10, it is the areas, not the heights, of rectangles that give the probabilities of the respec-
tive classes. Thus, if we add the areas of all the rectangles in Figure 6.10, we obtain the sum of all prob-
abilities equal to 1 .0001 , which is approximately 1 .0. Consequently, the total area under the curve is equal
to 1.0.
Class
lYCICHI VC rlcUUcI ILy
\J cl 1 jI Ly
20 to less than
25
.00058
25 to less than
30
.00398
30 to less than
35
.01478
35 to less than
40
.02108
40 to less than
45
.02180
45 to less than
50
.03182
50 to less than
55
.03358
55 to less than
60
.02580
60 to less than
65
.01534
65 to less than
70
.00804
70 to less than
75
.00600
75 to less than
80
.00456
80 to less than
85
.00342
85 to less than
90
.00342
90 to less than
95
.00286
95 to less than
100
.00206
100 to less than
105
.00044
105 to less than
110
.00030
1 10 to less than 115
.00016
The probability distribution of a continuous random variable has a mean and a standard deviation, denoted
by /it and a, respectively. The mean and standard deviation of the probability distribution curve of Figure
6.10 are 52.5401 and 14.7972 minutes, respectively. These values of fj. and a are calculated by using the
raw data on 10,431 participants.
Source: Data courtesy of the Manchester Road Race and Granite State Race Services.
256
6.2 The Normal Distribution 257
The normal probability distribution or the normal curve is a bell-shaped (symmetric)
curve. Such a curve is shown in Figure 6.1 1. Its mean is denoted by /jl and its standard deviation
by <j. A continuous random variable x that has a normal distribution is called a normal random
variable. Note that not all bell-shaped curves represent a normal distribution curve. Only a spe-
cific kind of bell-shaped curve represents a normal curve.
Standard
\^ deviation = a
Mean = n
Figure 6.T 1 Normal distribution with mean /jl and
standard deviation a.
Normal Probability Distribution A normal probability distribution, when plotted, gives a bell-
shaped curve such that:
1. The total area under the curve is 1.0.
2. The curve is symmetric about the mean.
3. The two tails of the curve extend indefinitely.
A normal distribution possesses the following three characteristics:
1. The total area under a normal distribution curve is 1.0, or 100%, as shown in Figure 6.12.
2. A normal distribution curve is symmetric about the mean, as shown in Figure 6.13. Con-
sequently, 50% of the total area under a normal distribution curve lies on the left side of
the mean, and 50% lies on the right side of the mean.
— Each of the two —
shaded areas
is .5 or 50%
.5
.5
Figure 6.13 A normal curve is symmetric about the mean.
The tails of a normal distribution curve extend indefinitely in both directions without touch-
ing or crossing the horizontal axis. Although a normal distribution curve never meets the
horizontal axis, beyond the points represented by /jl — 3<x and /jl + 3cr it becomes so close
to this axis that the area under the curve beyond these points in both directions can be taken
as virtually zero. These areas are shown in Figure 6.14.
258 Chapter 6 Continuous Random Variables and the Normal Distribution
The mean, fju, and the standard deviation, er, are the parameters of the normal distribution.
Given the values of these two parameters, we can find the area under a normal distribution curve
for any interval. Remember, there is not just one normal distribution curve but a family of nor-
mal distribution curves. Each different set of values of fi and a gives a different normal distri-
bution. The value of fi determines the center of a normal distribution curve on the horizontal
axis, and the value of a gives the spread of the normal distribution curve. The three normal dis-
tribution curves drawn in Figure 6.15 have the same mean but different standard deviations. By
contrast, the three normal distribution curves in Figure 6.16 have different means but the same
standard deviation.
Figure 6.15 Three normal distribution curves with the
same mean but different standard deviations.
\* ct-5
->A-< <7= 10
ct= 16
\i = 50 x
Figure 6.16 Three normal distribution curves with
different means but the same standard deviation.
fi = 20 n = 30 n = 40 x
Like the binomial and Poisson probability distributions discussed in Chapter 5, the normal
probability distribution can also be expressed by a mathematical equation. 1 However, we will
not use this equation to find the area under a normal distribution curve. Instead, we will use
Table IV of Appendix C.
'The equation of the normal distribution is
ff x \ = 1 g-imi'-n)/"]'
<tV2i7
where e = 2.71828 and tt = 3.14159 approximately; /(a), called the probability density function, gives the vertical distance
between the horizontal axis and the curve at point x. For the information of those who are familiar with integral calculus,
the definite integral of this equation from a to b gives the probability that x assumes a value between a and b.
6.3 The Standard Normal Distribution 259
6.3 The Standard Normal Distribution
The standard normal distribution is a special case of the normal distribution. For the stan-
dard normal distribution, the value of the mean is equal to zero, and the value of the standard
deviation is equal to 1.
Definition
Standard Normal Distribution The normal distribution with fi = and a = 1 is called the stan-
dard normal distribution.
Figure 6.17 displays the standard normal distribution curve. The random variable that
possesses the standard normal distribution is denoted by z. In other words, the units for the
standard normal distribution curve are denoted by z and are called the z values or z scores.
They are also called standard units or standard scores.
Definition
1 Values or 1 Scores The units marked on the horizontal axis of the standard normal curve are
denoted by z and are called the z values or z scores. A specific value of z gives the distance
between the mean and the point represented by z in terms of the standard deviation.
In Figure 6.17, the horizontal axis is labeled z. The z values on the right side of the mean are
positive and those on the left side are negative. The z value for a point on the horizontal axis gives
the distance between the mean and that point in terms of the standard deviation. For example, a
point with a value of z = 2 is two standard deviations to the right of the mean. Similarly, a point
with a value of z = —2 is two standard deviations to the left of the mean.
The standard normal distribution table, Table IV of Appendix C, lists the areas under the
standard normal curve to the left of z values from -3.49 to 3.49. To read the standard normal
distribution table, we look for the given z value in the table and record the value corresponding
to that z value. As shown in Figure 6.18, Table IV gives what is called the cumulative proba-
bility to the left of any z value.
Figure 6.18 Area under the standard normal curve.
260 Chapter 6 Continuous Random Variables and the Normal Distribution
Although the values of z on the left side of the mean are negative, the area under the curve is
always positive.
The area under the standard normal curve between any two points can be interpreted as the
probability that z assumes a value within that interval. Examples 6-1 through 6-4 describe how
to read Table IV of Appendix C to find areas under the standard normal curve.
■ EXAMPLE 6-1
Find the area under the standard normal curve to the left of z = 1 .95.
Solution We divide the given number 1.95 into two portions: 1.9 (the digit before the dec-
imal and one digit after the decimal) and .05 (the second digit after the decimal). (Note that
1.95 = 1.9 + .05.) To find the required area under the standard normal curve, we locate 1.9
in the column for z on the left side of Table IV and .05 in the row for z at the top of Table IV.
The entry where the row for 1.9 and the column for .05 intersect gives the area under the stan-
dard normal curve to the left of z = 1.95. The relevant portion of Table IV is reproduced as
Table 6.2. From Table IV or Table 6.2, the entry where the row for 1.9 and the column for .05
cross is .9744. Consequently, the area under the standard normal curve to the left of z = 1.95
is .9144. This area is shown in Figure 6.19. (It is always helpful to sketch the curve and mark
the area you are determining.)
Table 6.2 Area Under the Standard Normal Curve to the Left of z = 1.95
z
.00
.01
.09
-3.4
.0003
.0003
.0003
.0002
-3.3
.0005
.0005
.0004
.0003
-3.2
.0007
.0007
.0006
.0005
m
.9713
.9719
|.9744|^-
.9767
3.4
.9997
.9997
.9997
.9998
Required area
1.95 z
Figure 6.19 Area to the left of z = 1.95.
Remember ►
Finding the area to
the left of a positive z.
6.3 The Standard Normal Distribution
The area to the left of z = 1-95 can be interpreted as the probability that z assumes a value
less than 1.95; that is,
Area to the left of 1.95 = P(z < 1.95) = .9744
As mentioned in Section 6.1, the probability that a continuous random variable assumes a
single value is zero. Therefore,
P(z = 1.95) =
Hence,
P(z < 1.95) = P(z < 1.95) = .9744 |
■ EXAMPLE 6-2
Find the area under the standard normal curve from z = —2.17 to z = 0.
Finding the area
Solution To find the area from z = -2. 17 to z = 0, first we find the areas to the left of between a negative z and z
z = and to the left of z = —2.17 in the standard normal distribution table (Table IV). As
shown in Table 6.3, these two areas are .5 and .0150, respectively. Next we subtract .0150
from .5 to find the required area.
Table 6.3 Area Under the Standard Normal Curve
-2.17
The area from z = —2.17 to z = gives the probability that z lies in the interval —2.17 to
(see Figure 6.20); that is,
Area from -2.17 to = P(-2.17 < z < 0)
= P( z < 0) - P(z < -2.17) = .5000 - .0150 = .4850
262 Chapter 6 Continuous Random Variables and the Normal Distribution
■ EXAMPLE 6-3
Find the following areas under the standard normal curve.
(a) Area to the right of z = 2.32
(b) Area to the left of z = - 1.54
Solution
(a) As mentioned earlier, the normal distribution table gives the area to the left of a z
value. To find the area to the right of z = 2.32, first we find the area to the left of
z = 2.32. Then we subtract this area from 1.0, which is the total area under the curve.
From Table IV, the area to the left of z = 2.32 is .9898. Consequently, the required
area is 1.0 — .9898 = .0102, as shown in Figure 6.21.
Figure 6.21 Area to the right of z = 2.32.
. Shaded area
\ is 1.0 -.9898
/ .9898
1
N. = .0102
2.32
The area to the right of z = 2.32 gives the probability that z is greater than 2.32. Thus,
Area to the right of 2.32 = P(z > 2.32) = 1.0 - .9898 = .0102
(b) To find the area under the standard normal curve to the left of z = —1.54, we find
the area in Table IV that corresponds to —1.5 in the z column and .04 in the top row.
This area is .0618. This area is shown in Figure 6.22.
Figure 6.22 Area to the left of z = - 1.54.
-1 .54
The area to the left of z = — 1 .54 gives the probability that z is less than — 1 .54. Thus,
Area to the left of - 1 .54 = P(z < - 1 .54) = .0618 ■
■ EXAMPLE 6-4
Find the following probabilities for the standard normal curve.
(a) P(1.19 < z < 2.12) (b) P(-1.56 < z < 2.31) (c) P{z > -.75)
Solution
(a) The probability P(1.19 < z < 2.12) is given by the area under the standard normal
curve between z = 1.19 and z = 2.12, which is the shaded area in Figure 6.23.
To find the area between z = 1.19 and z = 2.12, first we find the areas to the left
of z = 1.19 and z = 2.12. Then we subtract the smaller area (to the left of z = 1.19)
from the larger area (to the left of z = 2.12).
Finding the areas in
the right and left tails.
Finding the
area between two positive
values of z-
6.3 The Standard Normal Distribution 263
1.19 2.12
Figure 6.23 Finding P( 1.1 9 < z < 2.12).
From Table IV for the standard normal distribution, we find
Area to the left of 1.19 = .8830
Area to the left of 2. 12 = .9830
Then, the required probability is
P(1.19 < z < 2.12) = Area between 1.19 and 2.12
= .9830 - .8830 = .1000
The probability P(— 1.56 < z < 2.31) is given by the area under the standard nor
mal curve between z = —1.56 and z = 2.31, which is the shaded area in Figure 6.24
From Table IV for the standard normal distribution, we have
Area to the left of -1.56 = .0594
Area to the left of 2.31 = .9896
The required probability is
P(-1.56 < z < 2.31) = Area between -1.56 and 2.31
= .9896 - .0594 = .9302
Finding the area
between a positive and a
negative value of z-
The probability P(z > —.75) is given by the area under the standard normal curve to
• r nc , ■ , ■ , , , 7—1- r- Finding the area to the
the right of z = —.75, which is the shaded area in Figure 6.25.
right of a negative value of z-
264 Chapter 6 Continuous Random Variables and the Normal Distribution
From Table IV for the standard normal distribution.
Area to the left of -.75 = .2266
The required probability is
P(z > -.75) = Area to the right of -.75 = 1.0 - .2266 = .7734 H
In the discussion in Section 3.4 of Chapter 3 on the use of the standard deviation, we discussed
the empirical rule for a bell-shaped curve. That empirical rule is based on the standard normal dis-
tribution. By using the normal distribution table, we can now verify the empirical rule as follows.
1. The total area within one standard deviation of the mean is 68.26%. This area is given by
the difference between the area to the left of z = 1.0 and the area to the left of z = —1.0.
As shown in Figure 6.26, this area is .8413 - .1587 = .6826, or 68.26%.
2. The total area within two standard deviations of the mean is 95.44%. This area is given by
the difference between the area to the left of z = 2.0 and the area to the left of z = —2.0.
As shown in Figure 6.27, this area is .9772 - .0228 = .9544, or 95.44%.
3. The total area within three standard deviations of the mean is 99.74%. This area is given
by the difference between the area to the left of z = 3.0 and the area to the left of z = —3.0.
As shown in Figure 6.28, this area is .9987 - .0013 = .9974, or 99.74%.
Again, note that only a specific bell-shaped curve represents the normal distribution. Now
we can state that a bell-shaped curve that contains (about) 68.26% of the total area within one
standard deviation of the mean, (about) 95.44% of the total area within two standard deviations
of the mean, and (about) 99.74% of the total area within three standard deviations of the mean
represents a normal distribution curve.
The standard normal distribution table, Table IV of Appendix C, goes from z = —3.49 to
z = 3.49. Consequently, if we need to find the area to the left of z = —3.50 or a smaller value of
Z, we can assume it to be approximately 0.0. If we need to find the area to the left of z = 3.50 or
a larger number, we can assume it to be approximately 1 .0. Example 6-5 illustrates this procedure.
6.3 The Standard Normal Distribution 265
■ EXAMPLE 6-5
Find the following probabilities for the standard normal curve,
(a) P(0 < z < 5.67) (b) P(z < -5.35)
Solution
(a) The probability P(Q < z < 5.67) is given by the area under the standard normal
curve between z = and z = 5.67. Because z = 5.67 is greater than 3.49 and is not
in Table IV, the area under the standard normal curve to the left of z = 5.67 can be
approximated by 1.0. Also, the area to the left of z = is .5. Hence, the required
probability is
P(0 < z < 5.67) = Area between and 5.67 = 1.0 — .5 = .5 approximately
Note that the area between z = and z = 5.67 is not exactly .5 but very close to .5.
This area is shown in Figure 6.29.
Finding the area
between z = and a value of z
greater than 3.49.
(b) The probability P(z < —5.35) represents the area under the standard normal curve to
the left of z = —5.35. Since z = —5.35 is not in the table, we can assume that this
area is approximately .00. This is shown in Figure 6.30.
Finding the area to the left
of a z that is less than —3.49.
The required probability is
P{z < -5.35) = Area to the left of -5.35 = .00 approximately
Again, note that the area to the left of z = —5.35 is not exactly .00 but very close to .00.
We can find the exact areas for parts (a) and (b) of this example by using technol-
ogy. The reader should do that. g
EXERCISES
CONCEPTS AND PROCEDURES
6.1 What is the difference between the probability distribution of a discrete random variable and that of
a continuous random variable? Explain.
6.2 Let x be a continuous random variable. What is the probability that x assumes a single value, such as at
6.3 For a continuous probability distribution, why is P(a < x < b) equal to P(a £ x < b)l
6.4 Briefly explain the main characteristics of a normal distribution. Illustrate with the help of graphs.
6.5 Briefly describe the standard normal distribution curve.
266 Chapter 6 Continuous Random Variables and the Normal Distribution
6.6 What are the parameters of the normal distribution?
6.7 How do the width and height of a normal distribution change when its mean remains the same but
its standard deviation decreases?
6.8 Do the width and/or height of a normal distribution change when its standard deviation remains the
same but its mean increases?
6.9 For the standard normal distribution, what does z represent?
6.10 For the standard normal distribution, find the area within one standard deviation of the mean — that
is, the area between jju — cr and fi + a.
6.11 For the standard normal distribution, find the area within 1.5 standard deviations of the mean — that
is, the area between /jl, — 1.5cr and ju + 1.5o\
6.12 For the standard normal distribution, what is the area within two standard deviations of the mean?
6.13 For the standard normal distribution, what is the area within 2.5 standard deviations of the mean?
6.14 For the standard normal distribution, what is the area within three standard deviations of the mean?
6.15 Find the area under the standard normal curve
a. between z = and z = 1.95 b. between z = and z = —2.05
c. between z = 1.15 and z = 2.37 d. from z = -1.53 to z = -2.88
e. from z = -1.67 to z = 2.24
6.16 Find the area under the standard normal curve
a. from z = to z = 2.34 b. between z
c. from z = .84 to z = 1.95 d. between z
e. between z = —2.15 and z = 1.87
6.17 Find the area under the standard normal curve
a. to the right of z = 1.36 b. to the left of z = -1.97
c. to the right of z = -2.05 d. to the left of z = 1.76
6.18 Obtain the area under the standard normal curve
a. to the right of z = 1.43 b. to the left of z = -1.65
c. to the right of z = -.65 d. to the left of z = .89
6.19 Find the area under the standard normal curve
a. between z = and z = 4.28 b. from z = to z = —3.75
c. to the right of z = 7.43 d. to the left of z = -4.69
6.20 Find the area under the standard normal curve
a. from z = to z = 3.94 b. between z = and z = —5.16
c. to the right of z = 5.42 d. to the left of z = -3.68
6.21 Determine the following probabilities for the standard normal distribution,
a. P(-1.83 <z< 2.57) b. P(0 < z < 2.02)
c. P(-1.99 < z < 0) d. P(z>1.48)
6.22 Determine the following probabilities for the standard normal distribution,
a. P(-2.46 < z < 1.88) b. P(0 < z < 1.96)
c. P(-2.58 < z < 0) d. P(z > .73)
6.23 Find the following probabilities for the standard normal distribution,
a. P(z < -2.34) b. P(.67 < z < 2.59)
c. P(-2.07 < z < -.93) d. P(z < 1.78)
6.24 Find the following probabilities for the standard normal distribution,
a. P(z< -1.31) b. P( 1.23 < z < 2.89)
c. P(-2.24 < z < -1.19) d. P(z < 2.02)
6.25 Obtain the following probabilities for the standard normal distribution,
a. P(z > -.98) b. P(-2.47 < z < 1.29)
c. P(0 < z < 4.25) d. P(-5.36 < z < 0)
e. P(z > 6.07) f. P(z < -5.27)
6.26 Obtain the following probabilities for the standard normal distribution,
a. P(z > -1.86) b. P(-.68 < z < 1.94)
c. P(0 < z < 3.85) d. P(-4.34 < z < 0)
e. P(z > 4.82) f. P(z < -6.12)
= and z = -2.58
= -.57 and z = -2.49
6.4 Standardizing a Normal Distribution 267
6.4 Standardizing a Normal Distribution
As was shown in the previous section, Table IV of Appendix C can be used to find areas un-
der the standard normal curve. However, in real-world applications, a (continuous) random vari-
able may have a normal distribution with values of the mean and standard deviation that are
different from and 1, respectively. The first step in such a case is to convert the given normal
distribution to the standard normal distribution. This procedure is called standardizing a nor-
mal distribution. The units of a normal distribution (which is not the standard normal distribu-
tion) are denoted by x. We know from Section 6.3 that units of the standard normal distribution
are denoted by z.
Converting an x Value to a z Value For a normal random variable x, a particular value of x can
be converted to its corresponding z value by using the formula
X — fJb
z =
cr
where /jl and cr are the mean and standard deviation of the normal distribution of x,
respectively.
Thus, to find the z value for an x value, we calculate the difference between the given x
value and the mean, /a, and divide this difference by the standard deviation, cr. If the value of
x is equal to fi, then its z value is equal to zero. Note that we will always round z values to two
decimal places.
The z value for the mean of a normal distribution is always zero.
Examples 6-6 through 6-10 describe how to convert x values to the corresponding z val-
ues and how to find areas under a normal distribution curve.
■ EXAMPLE 6-6
Let x be a continuous random variable that has a normal distribution with a mean of 50 and
Converting x values
a standard deviation of 10. Convert the following x values to z values and find the probabil- ,
D r to the corresponding z values.
ity to the left of these points,
(a) x = 55 (b) x = 35
Solution For the given normal distribution, fx = 50 and cr = 10.
(a) The z value for x = 55 is computed as follows:
x- ix 55-50
10
.50
Thus, the z value for x = 55 is .50. The z values for /jl = 50 and x = 55 are
shown in Figure 6.31. Note that the z value for fx = 50 is zero. The value z = .50
for x = 55 indicates that the distance between /a = 50 and x = 55 is 1/2 of the
standard deviation, which is 10. Consequently, we can state that the z value rep-
resents the distance between /a and x in terms of the standard deviation. Because
jc = 55 is greater than jx = 50, its z value is positive.
268 Chapter 6 Continuous Random Variables and the Normal Distribution
Figure 6.31 z value forx = 55.
From this point on, we will usually show only the z axis below the x axis and not
the standard normal curve itself.
To find the probability to the left of x = 55, we find the probability to the left of
z = .50 from Table IV. This probability is .6915. Therefore,
P(x < 55) = P(z < .50) = .6915
(b) The z value for x = 35 is computed as follows and is shown in Figure 6.32:
z = ^ = 35-50
er 10
Because x = 35 is on the left side of the mean (i.e., 35 is less than jj, = 50), its z
value is negative. As a general rule, whenever an x value is less than the value of /x,
its z value is negative.
To find the probability to the left of x = 35, we find the area under the normal
curve to the left of z = —1.50. This area from Table IV is .0668. Hence,
P{x < 35) = P(z < -1.50) = .0668
Remember ► The z value for an x value that is greater than /jl is positive, the z value for an x value that is
equal to /jl is zero, and the z value for an x value that is less than fi is negative.
To find the area between two values of x for a normal distribution, we first convert both
values of x to their respective z values. Then we find the area under the standard normal curve
6.4 Standardizing a Normal Distribution 269
between those two z values. The area between the two z values gives the area between the cor-
responding x values. Example 6-7 illustrates this case.
■ EXAMPLE 6-7
Let x be a continuous random variable that is normally distributed with a mean of 25 and a
standard deviation of 4. Find the area
(a) between x = 25 and x = 32 (b) between x = 18 and x = 34
Solution For the given normal distribution, /j, = 25 and cr = 4.
(a) The first step in finding the required area is to standardize the given normal distribu-
tion by converting x = 25 and x = 32 to their respective z values using the formula
x — i±
z =
(T
The z value for x = 25 is zero because it is the mean of the normal distribution. The
z value for x = 32 is
= 1.75
The area between x = 25 and x = 32 under the given normal distribution curve is equi-
valent to the area between z = and z = 1.75 under the standard normal curve. From
Table IV, the area to the left of z = 1.75 is .9599, and the area to the left of z = is
.50. Hence, the required area is .9599 — .50 = .4599, which is shown in Figure 6.33.
The area between x = 25 and x = 32 under the normal curve gives the probabil-
ity that x assumes a value between 25 and 32. This probability can be written as
P(25 < x < 32) = P(Q < z < 1.75) = .4599
Finding the area
between the mean and a
point to its right.
(b) First, we calculate the z values for x
Forx = 18:
Forx = 34:
18 and x = 34 as follows:
18 - 25
= -1.75
4
34 - 25
Finding the area
between two points on different
sides of the mean
2.25
The area under the given normal distribution curve between x = 18 and x = 34 is
given by the area under the standard normal curve between z = —1.75 and z = 2.25.
From Table IV, the area to the left of z = 2.25 is .9878, and the area to the left of
z = —1.75 is .0401. Hence, the required area is
P(18 < x < 34) = P(-1.75 < z < 2.25) = .9878 - .0401 = .9477
This area is shown in Figure 6.34.
270 Chapter 6 Continuous Random Variables and the Normal Distribution
■ EXAMPLE 6-8
Let x be a normal random variable with its mean equal to 40 and standard deviation equal to
5. Find the following probabilities for this normal distribution.
(a) P(x > 55) (b) P(x < 49)
Solution For the given normal distribution, /jl = 40 and cr = 5.
(a) The probability that x assumes a value greater than 55 is given by the area under the
normal distribution curve to the right of x = 55, as shown in Figure 6.35. This area
is calculated by subtracting the area to the left of x = 55 from 1.0, which is the total
area under the curve.
55 - 40
Forx = 55: z = = 3.00
Calculating the probability
of x falling in the right tail.
Calculating the probability
that x is less than a value to the
right of the mean.
The required probability is given by the area to the right of z = 3.00. To find this area,
first we find the area to the left of z — 3.00, which is .9987. Then we subtract this
area from 1.0. Thus,
P{x > 55) = P(z > 3.00) = 1.0 - .9987 = .0013
(b) The probability that x will assume a value less than 49 is given by the area under the
normal distribution curve to the left of 49, which is the shaded area in Figure 6.36.
This area is obtained from Table IV.
Forx = 49: z
49 - 40
1.80
6.4 Standardizing a Normal Distribution 271
The required probability is given by the area to the left of z — 1-80. This area from
Table IV is .9641. Therefore, the required probability is
P(x < 49) = P(z < 1.80) = .9641 ■
■ EXAMPLE 6-9
Let x be a continuous random variable that has a normal distribution with /j, = 50 and
o- = 8. Find the probability P(30 < x < 39).
Solution For this normal distribution, i± = 50 and cr = 8. The probability P(30 < x < 39) is
given by the area from x = 30 to x = 39 under the normal distribution curve. As shown in Fig-
ure 6.37, this area is given by the difference between the area to the left of x = 30 and the area
to the left ofx = 39.
For* = 30: z = =-2.50
Figure 6.37 Finding P(30 < x < 39).
Finding the area between
two x values that are less
than the mean.
To find the required area, we first find the area to the left of z = —2.50, which is .0062.
Then, we find the area to the left of z = —1.38, which is .0838. The difference between these
two areas gives the required probability, which is
P(30 < x < 39) = P(-2.50 < z < -1.38) = .0838 - .0062 = .0776 ■
■ EXAMPLE 6-10
Let x be a continuous random variable that has a normal distribution with a mean of 80 and
a standard deviation of 12. Find the area under the normal distribution curve
(a) from x = 70 to x = 135 (b) to the left of 27
Solution For the given normal distribution, /a = 80 and cr = 12.
(a) The z values for x = 70 and x = 135 are:
70 - 80
Forx = 70: z
12
.83
Finding the area between
two x values that are on different
sides of the mean.
135 - 80
Forx =135: z = = 4.58
12
Thus, to find the required area we find the areas to the left of z = — .83 and to the left
of z = 4.58 under the standard normal curve. From Table IV, the area to the left of
272 Chapter 6 Continuous Random Variables and the Normal Distribution
Figure 6.38 Area between x = 70 and x = 135.
>y Shaded area
\ is approximately
\ =.7967
I r-
70 80 135 x
58
z = — .83 is .2033 and the area to the left of z = 4.58 is approximately 1.0. Note that
z = 4.58 is not in Table IV.
Hence,
P(70 < x < 135) = P(-.83 < z < 4.58)
= 1.0 - .2033 = .7967 approximately
Figure 6.38 shows this area,
(b) First we find the z value for x = 27.
27 - 80
Forjc = 27: z = = -4.42
12
As shown in Figure 6.39, the required area is given by the area under the standard
normal distribution curve to the left of z = —4.42. This area is approximately zero.
P(x < 27) = P(z < -4.42) = .00 approximately
EXERCISES
CONCEPTS AND PROCEDURES
6.27 Find the z value for each of the following x values for a normal distribution with (i = 30 and
a = 5.
a. x = 39 b. x — 19 c. x = 24 d. x = 44
6.28 Determine the z value for each of the following x values for a normal distribution with /x
(7=3.
a. x = 12 b. x = 22 c. x = 19 d. x = 13
6.29 Find the following areas under a normal distribution curve with /jl = 20 and a = 4.
a. Area between x = 20 and x = 27
b. Area from x = 23 to x — 26
c. Area between x = 9.5 and x = 17
6.30 Find the following areas under a normal distribution curve with fi = 12 and a = 2.
a. Area between x = 7.76 and x = 12
b. Area between x = 14.48 and x = 16.54
c. Area from x = 8.22 to x = 10.06
16 and
6.5 Applications of the Normal Distribution 273
6.31 Determine the area under a normal distribution curve with jx = 55 and cr = 1
a. to the right of x = 58 b. to the right of x = 43
c. to the left of x = 68 d. to the left of x = 22
6.32 Find the area under a normal distribution curve with /a = 37 and cr = 3
a. to the left of x = 30 b. to the right of x = 52
c. to the left of x = 44 d. to the right of x = 32
6.33 Let x be a continuous random variable that is normally distributed with a mean of 25 and a standard
deviation of 6. Find the probability that x assumes a value
a. between 29 and 36 b. between 22 and 35
6.34 Let x be a continuous random variable that has a normal distribution with a mean of 40 and a stan-
dard deviation of 4. Find the probability that x assumes a value
a. between 29 and 35 b. from 34 to 50
6.35 Let x be a continuous random variable that is normally distributed with a mean of 80 and a standard
deviation of 12. Find the probability that x assumes a value
a. greater than 69 b. less than 73
c. greater than 101 d. less than 87
6.36 Let x be a continuous random variable that is normally distributed with a mean of 65 and a standard
deviation of 15. Find the probability that x assumes a value
a. less than 45 b. greater than 79
c. greater than 54 d. less than 70
6.5 Applications of the Normal Distribution
Sections 6.2 through 6.4 discussed the normal distribution, how to convert a normal distribu-
tion to the standard normal distribution, and how to find areas under a normal distribution
curve. This section presents examples that illustrate the applications of the normal distribution.
■ EXAMPLE 6-11
According to a Sallie Mae survey and credit bureau data, in 2008, college students carried an
average of $3173 debt on their credit cards (USA TODAY, April 13, 2009). Suppose that cur-
rent credit card debts for all college students have a normal distribution with a mean of $3173
and a standard deviation of $800. Find the probability that credit card debt for a randomly se-
lected college student is between $2109 and $3605.
Solution Let x denote the credit card debt of a randomly selected college student. Then, x
is normally distributed with
/X = $3173 and cr = $800
The probability that the credit card debt of a randomly selected college student is between
$2109 and $3605 is given by the area under the normal distribution curve of x that falls be-
tween x = $2109 and x = $3605, as shown in Figure 6.40. To find this area, first we find the
areas to the left of x = $2109 and x = $3605, respectively, and then take the difference be-
tween these two areas.
2109 - 3173
Forx = $2109: z = — =-1.33
800
3605 - 3173
For x = $3605: z = = .54
800
Thus, the required probability is given by the difference between the areas under the standard
normal curve to the left of z = —1.33 and to the left of z =.54. From Table IV in Appendix C,
Using the normal distribution:
the area between two points on
different sides of the mean.
274 Chapter 6 Continuous Random Variables and the Normal Distribution
Figure 6.40 Area between x = $2109 and
x = $3605.
Shaded area
is .6136
$2109
$3173 $3605
-1.33
1~
.54
the area to the left of z
required probability is
1.33 is .0918, and the area to the left of z = .54 is .7054. Hence, the
P($2109 < x < $3605) = P(-1.33 < z < .54) = .7054 - .0918
.6136
Thus, the probability is .6136 that the credit card debt of a randomly selected college stu-
dent is between $2109 and $3605. Converting this probability into a percentage, we can also
state that (about) 61.36% of all college students have credit card debts between $2109 and
$3605. ■
Using the normal
distribution: probability that
x is less than a value that is
to the right of the mean.
■ EXAMPLE 6-12
A racing car is one of the many toys manufactured by Mack Corporation. The assembly times
for this toy follow a normal distribution with a mean of 55 minutes and a standard deviation
of 4 minutes. The company closes at 5 p.m. every day. If one worker starts to assemble a rac-
ing car at 4 p.m., what is the probability that she will finish this job before the company closes
for the day?
Solution Let x denote the time this worker takes to assemble a racing car. Then, x is nor-
mally distributed with
fx — 55 minutes and cr = 4 minutes
We are to find the probability that this worker can assemble this car in 60 minutes or less
(between 4 and 5 p.m.). This probability is given by the area under the normal curve to the
left of x = 60 as shown in Figure 6.41.
Forx = 60: z = =1.25
4
The required probability is given by the area under the standard normal curve to the left of
z = 1.25, which is .8944 from Table IV of Appendix C. Thus, the required probability is
P(x < 60) = P(z < 1.25) = .8944
Thus, the probability is .8944 that this worker will finish assembling this racing car before the
company closes for the day. H
6.5 Applications of the Normal Distribution 275
■ EXAMPLE 6-13
Hupper Corporation produces many types of soft drinks, including Orange Cola. The filling
machines are adjusted to pour 12 ounces of soda into each 12-ounce can of Orange Cola. How-
ever, the actual amount of soda poured into each can is not exactly 12 ounces; it varies from
can to can. It has been observed that the net amount of soda in such a can has a normal dis-
tribution with a mean of 12 ounces and a standard deviation of .015 ounce.
(a) What is the probability that a randomly selected can of Orange Cola contains 11.97
to 11.99 ounces of soda?
(b) What percentage of the Orange Cola cans contain 12.02 to 12.07 ounces of soda?
Using the normal distribution.
Solution Let x be the net amount of soda in a can of Orange Cola. Then, x has a normal
distribution with /jl = 12 ounces and a = .015 ounce.
(a) The probability that a randomly selected can contains 11.97 to 11.99 ounces of soda
is given by the area under the normal distribution curve from x = 1 1.97 to x = 1 1.99.
This area is shown in Figure 6.42.
Forx = 11.97:
Forx = 11.99:
11.97 - 12
.015
11.99 - 12
.015
-2.00
-.67
Calculating the probability
between two points that are
to the left of the mean.
Shaded ar
= .251 4 -.02
= .22
ea — i >
28 /
86 /
1
11.97 11.9912
X
i — r
>7
Figure 6.42 Area between x = 11.97 and x = 11.99.
(b)
The required probability is given by the area under the standard normal curve be-
tween z = —2.00 and z = —.67. From Table IV of Appendix C, the area to the left
of z = -2.00 is .0228, and the area to the left of z = -.67 is .2514. Hence, the
required probability is
P (11.97
11.99) = P(-2.00
.67)
.2514 - .0228
.2286
Thus, the probability is .2286 that any randomly selected can of Orange Cola will
contain 11.97 to 11.99 ounces of soda. We can also state that about 22.86% of Or-
ange Cola cans contain 11.97 to 11.99 ounces of soda.
The percentage of Orange Cola cans that contain 12.02 to 12.07 ounces of soda is
given by the area under the normal distribution curve from x = 12.02 to x = 12.07,
as shown in Figure 6.43.
12.02 - 12
Forx= 12.02: z = — =1.33
12.07 - 12
For x = 12.07: z = — = 4.67
Calculating the probability
between two points that are to
the right of the mean.
276 Chapter 6 Continuous Random Variables and the Normal Distribution
Figure 6.43 Area from x = 12.02 to x = 12.07.
H= 12 12.02
12.07
1.33
4.67
The required probability is given by the area under the standard normal curve between
z = 1.33 and z = 4.67. From Table IV of Appendix C, the area to the left of z = 1-33
is .9082, and the area to the left of z = 4.67 is approximately 1 .0. Hence, the required
probability is
P(12.02 < x < 12.07) = P (1.33 < z < 4.67) = 1.0 - .9082
.0918
Converting this probability to a percentage, we can state that approximately 9.18%
of all Orange Cola cans are expected to contain 12.02 to 12.07 ounces of soda. I
EXAMPLE 6-14
Finding the area to the left
of x that is less than the mean.
Suppose the life span of a calculator manufactured by Texas Instruments has a normal distribution
with a mean of 54 months and a standard deviation of 8 months. The company guarantees that any
calculator that starts malfunctioning within 36 months of the purchase will be replaced by a new
one. About what percentage of calculators made by this company are expected to be replaced?
Solution Let x be the life span of such a calculator. Then x has a normal distribution with
fi = 54 and a = 8 months. The probability that a randomly selected calculator will start to
malfunction within 36 months is given by the area under the normal distribution curve to the
left of x = 36, as shown in Figure 6.44.
Forjc = 36: z
36 - 54
-2.25
Shaded area
is .0122
p = 54
-2.25
Figure 6.44 Area to the left of x = 36.
The required percentage is given by the area under the standard normal curve to the left of
z = —2.25. From Table IV of Appendix C, this area is .0122. Hence, the required probability is
P{x < 36) = P(z < -2.25) = .0122
The probability that any randomly selected calculator manufactured by Texas Instruments
will start to malfunction within 36 months is .0122. Converting this probability to a percent-
age, we can state that approximately 1.22% of all calculators manufactured by this company
are expected to start malfunctioning within 36 months. Hence, 1.22% of the calculators are
expected to be replaced. H
6.5 Applications of the Normal Distribution
277
EXERCISES
APPLICATIONS
6.37 Let x denote the time taken to run a road race. Suppose x is approximately normally distributed with
a mean of 190 minutes and a standard deviation of 21 minutes. If one runner is selected at random, what
is the probability that this runner will complete this road race
6.38 According to the 2007 American Time Use Survey by the Bureau of Labor Statistics, employed adults
living in households with no children younger than 18 years engaged in leisure activities for 4.4 hours a
day on average {Source: http://www.bls.gov/news.release/atus.nrO.htm). Assume that currently such times
are (approximately) normally distributed with a mean of 4.4 hours per day and a standard deviation of
1.08 hours per day. Find the probability that the amount of time spent on leisure activities per day for a
randomly chosen individual from the population of interest (employed adults living in households with no
children younger than 18 years) is
a. between 3.0 and 5.0 hours per day b. less than 2.0 hours per day
6.39 In a 2007 survey of consumer spending habits, U.S. residents aged 45 to 54 years spent an average
of 9.32% of their after-tax income on food (Source: ftp://ftp.bls.gov/pub/special.requests/ce/standard/2007/
age.txt). Suppose that the current percentage of after-tax income spent on food by all U.S. residents aged
45 to 54 years follows a normal distribution with a mean of 9.32% and a standard deviation of 1.38%.
Find the proportion of such persons whose percentage of after-tax income spent on food is
a. greater than 11.1% b. between 6.0% and 7.2%
6.40 Tommy Wait, a minor league baseball pitcher, is notorious for taking an excessive amount of time
between pitches. In fact, his times between pitches are normally distributed with a mean of 36 seconds
and a standard deviation of 2.5 seconds. What percentage of his times between pitches are
a. longer than 39 seconds? b. between 29 and 34 seconds?
6.41 A construction zone on a highway has a posted speed limit of 40 miles per hour. The speeds of ve-
hicles passing through this construction zone are normally distributed with a mean of 46 miles per hour
and a standard deviation of 4 miles per hour. Find the percentage of vehicles passing through this con-
struction zone that are
a. exceeding the posted speed limit
b. traveling at speeds between 50 and 57 miles per hour
6.42 The Bank of Connecticut issues Visa and MasterCard credit cards. It is estimated that the bal-
ances on all Visa credit cards issued by the Bank of Connecticut have a mean of $845 and a standard
deviation of $270. Assume that the balances on all these Visa cards follow a normal distribution.
a. What is the probability that a randomly selected Visa card issued by this bank has a balance be-
tween $1000 and $1440?
b. What percentage of the Visa cards issued by this bank have a balance of $730 or more?
6.43 According to an article published on the Web site www.PCMag.com, Facebook users spend an av-
erage of 190 minutes per month checking and updating their Facebook pages (Source: http://www.pcmag.com/
article2/0,2817,2342757,00.asp). Suppose that the current distribution of time spent per month checking
and updating a member's Facebook page is normally distributed with a mean of 190 minutes and a stan-
dard deviation of 53.4 minutes. For a randomly selected Facebook member, determine the probability that
the amount of time that he or she spends per month checking and updating his or her Facebook page is
a. more than 300 minutes b. between 120 and 180 minutes
6.44 The transmission on a model of a specific car has a warranty for 40,000 miles. It is known that the
life of such a transmission has a normal distribution with a mean of 72,000 miles and a standard devia-
tion of 13,000 miles.
a. What percentage of the transmissions will fail before the end of the warranty period?
b. What percentage of the transmissions will be good for more than 100,000 miles?
6.45 According to the records of an electric company serving the Boston area, the mean electricity con-
sumption for all households during winter is 1650 kilowatt-hours per month. Assume that the monthly
electricity consumptions during winter by all households in this area have a normal distribution with a
mean of 1650 kilowatt-hours and a standard deviation of 320 kilowatt-hours.
a. Find the probability that the monthly electricity consumption during winter by a randomly se-
lected household from this area is less than 1950 kilowatt-hours.
b. What percentage of the households in this area have a monthly electricity consumption of 900 to
1300 kilowatt-hours?
a. in less than 160 minutes?
b. in 215 to 245 minutes?
278
Chapter 6 Continuous Random Variables and the Normal Distribution
6.46 The management of a supermarket wants to adopt a new promotional policy of giving a free gift
to every customer who spends more than a certain amount per visit at this supermarket. The expectation
of the management is that after this promotional policy is advertised, the expenditures for all customers
at this supermarket will be normally distributed with a mean of $95 and a standard deviation of $20. If
the management decides to give free gifts to all those customers who spend more than $130 at this su-
permarket during a visit, what percentage of the customers are expected to get free gifts?
6.47 One of the cars sold by Walt's car dealership is a very popular subcompact car called Rhino. The fi-
nal sale price of the basic model of this car varies from customer to customer depending on the negotiat-
ing skills and persistence of the customer. Assume that these sale prices of this car are normally distrib-
uted with a mean of $19,800 and a standard deviation of $350.
a. Dolores paid $19,445 for her Rhino. What percentage of Walt's customers paid less than Dolores
for a Rhino?
b. Cuthbert paid $20,300 for a Rhino. What percentage of Walt's customers paid more than Cuth-
bert for a Rhino?
6.48 A psychologist has devised a stress test for dental patients sitting in the waiting rooms. Accord-
ing to this test, the stress scores (on a scale of 1 to 10) for patients waiting for root canal treatments
are found to be approximately normally distributed with a mean of 7.59 and a standard deviation of .73.
a. What percentage of such patients have a stress score lower than 6.0?
b. What is the probability that a randomly selected root canal patient sitting in the waiting room has
a stress score between 7.0 and 8.0?
c. The psychologist suggests that any patient with a stress score of 9.0 or higher should be given a
sedative prior to treatment. What percentage of patients waiting for root canal treatments would
need a sedative if this suggestion is accepted?
6.49 According to a 2004 survey by the telecommunications division of British Gas {Source: http://www.
literacytrust.org.uk/Database/texting. html#quarter), Britons spend an average of 225 minutes per day com-
municating electronically (on a fixed landline phone, on a mobile phone, by emailing, by texting, and
so on). Assume that currently such times for all Britons are normally distributed with a mean of 225 min-
utes per day and a standard deviation of 62 minutes per day. What percentage of Britons communicate
electronically for
a. less than 60 minutes per day b. more than 360 minutes per day
c. between 120 and 180 minutes per day d. between 240 and 300 minutes per day?
6.50 Fast Auto Service guarantees that the maximum waiting time for its customers is 20 minutes for oil
and lube service on their cars. It also guarantees that any customer who has to wait longer than 20 minutes
for this service will receive a 50% discount on the charges. It is estimated that the mean time taken for
oil and lube service at this garage is 15 minutes per car and the standard deviation is 2.4 minutes. Sup-
pose the time taken for oil and lube service on a car follows a normal distribution.
a. What percentage of the customers will receive the 50% discount on their charges?
b. Is it possible that a car may take longer than 25 minutes for oil and lube service? Explain.
6.51 The lengths of 3-inch nails manufactured on a machine are normally distributed with a mean of 3.0
inches and a standard deviation of .009 inch. The nails that are either shorter than 2.98 inches or longer
than 3.02 inches are unusable. What percentage of all the nails produced by this machine are unusable?
6.52 The pucks used by the National Hockey League for ice hockey must weigh between 5.5 and 6.0 ounces.
Suppose the weights of pucks produced at a factory are normally distributed with a mean of 5.75 ounces
and a standard deviation of .11 ounce. What percentage of the pucks produced at this factory cannot be
used by the National Hockey League?
6.6 Determining the z and x Values When
an Area Under the Normal Distribution
Curve Is Known
So far in this chapter we have discussed how to find the area under a normal distribution curve
for an interval of z or x. Now we reverse this procedure and learn how to find the correspon-
ding value of z or x when an area under a normal distribution curve is known. Examples 6-15
through 6-17 describe this procedure for finding the z value.
6.6 Determining the z and x Values When an Area Under the Normal Distribution Curve Is Known 279
■ EXAMPLE 6-15
Find a point z such that the area under the standard normal curve to the left of z is .925 1 .
Solution As shown in Figure 6.45, we are to find the z value such that the area to the left
of z is .9251. Since this area is greater than .50, z is positive and lies to the right of zero.
Figure 6.45 Finding the z value
Finding z when the
area to the left of z is known.
Shaded area
is given to be .9251
■To find this z
To find the required value of z, we locate .925 1 in the body of the normal distribution table,
Table IV of Appendix C. The relevant portion of that table is reproduced as Table 6.4 here.
Next we read the numbers in the column and row for z that correspond to .9251. As shown in
Table 6.4, these numbers are 1.4 and .04, respectively. Combining these two numbers, we obtain
the required value of z — 1 -44.
Table 6.4 Finding the z Value When Area Is Known
z
.00
.01
m
.09
-3.4
.0003
.0003
.0002
-3.3
.0005
.0005
.0003
-3.2
.0007
.0007
.0005
l.4k-
3.4
.9997
.9997
.9251
.9997
.9998
We locate this
value in Table IV
of Appendix C
■ EXAMPLE 6-16
Find the value of z such that the area under the standard normal curve in the right tail is .0050
Solution To find the required value of z, we first find the area to the left of z. Hence,
Area to the left of z = 1.0 - .0050 = .9950
This area is shown in Figure 6.46.
Now we look for .9950 in the body of the normal distribution table. Table IV does not
contain .9950. So we find the value closest to .9950, which is either .9949 or .9951. We can
use either of these two values. If we choose .9951, the corresponding z value is 2.58. Hence,
the required value of z is 2.58, and the area to the right of z = 2.58 is approximately .0050.
Note that there is no apparent reason to choose .9951 and not to choose .9949. We can use
either of the two values. If we choose .9949, the corresponding z value will be 2.57.
Finding z when the
area in the right tail is known.
280 Chapter 6 Continuous Random Variables and the Normal Distribution
Finding z when the area in the
left tail is known.
■ EXAMPLE 6-17
Find the value of z such that the area under the standard normal curve in the left tail is .05.
Solution Because .05 is less than .5 and it is the area in the left tail, the value of z is negative.
This area is shown in Figure 6.47.
Next, we look for .0500 in the body of the normal distribution table. The value closest to
.0500 in the normal distribution table is either .0505 or .0495. Suppose we use the value .0495.
The corresponding z value is —1.65. Thus, the required value of z is —1.65 and the area to
the left of z — —1.65 is approximately .05. H
To find an x value when an area under a normal distribution curve is given, first we find
the z value corresponding to that x value from the normal distribution table. Then, to find
the x value, we substitute the values of fi, <x, and z in the following formula, which is ob-
tained from z = (x — fJi)/(r by doing some algebraic manipulations. Also, if we know the
values of x, z, and a, we can find fi using this same formula. Exercises 6.63 and 6.64 present
such cases.
Finding an x Value for a Normal Distribution For a normal curve, with known values of /jl and
<j and for a given area under the curve to the left of x, the x value is calculated as
x = fj, + zcr
Examples 6-18 and 6-19 illustrate how to find an x value when an area under a normal
distribution curve is known.
■ EXAMPLE 6-18
Recall Example 6-14. It is known that the life of a calculator manufactured by Texas Instru-
ments has a normal distribution with a mean of 54 months and a standard deviation of 8 months.
What should the warranty period be to replace a malfunctioning calculator if the company
does not want to replace more than 1% of all the calculators sold?
Solution Let x be the life of a calculator. Then, x follows a normal distribution with
jx = 54 months and cr = 8 months. The calculators that would be replaced are the ones that
Finding x when the
area in the left tail is known.
6.6 Determining the z and x Values When an Area Under the Normal Distribution Curve Is Known 281
start malfunctioning during the warranty period. The company's objective is to replace at most
1% of all the calculators sold. The shaded area in Figure 6.48 gives the proportion of calcu-
lators that are replaced. We are to find the value of x so that the area to the left of x under the
normal curve is 1%, or .01.
Figure 6.48 Finding an x value.
To find this x /j = 54 x
x value
-2.33
t
1 — From Table IV, this value
of z is approximately -2.33
In the first step, we find the z value that corresponds to the required x value.
We find the z value from the normal distribution table for .0100. Table IV of Appendix C
does not contain a value that is exactly .0100. The value closest to .0100 in the table is .0099,
and the z value for .0099 is —2.33. Hence,
z = -2.33
Substituting the values of /x, a, and z in the formula x = n + zcr, we obtain
x = fL + zcr = 54 + (-2.33)(8) = 54 - 18.64 = 35.36
Thus, the company should replace all the calculators that start to malfunction within 35.36
months (which can be rounded to 35 months) of the date of purchase so that they will not have
to replace more than 1 % of the calculators. H
■ EXAMPLE 6-19
Almost all high school students who intend to go to college take the SAT test. In a recent
test, the mean SAT score (in verbal and mathematics) of all students was 1020. Debbie is
planning to take this test soon. Suppose the SAT scores of all students who take this test with
Debbie will have a normal distribution with a mean of 1020 and a standard deviation of 153.
What should her score be on this test so that only 10% of all examinees score higher than
she does?
Solution Let x represent the SAT scores of examinees. Then, x follows a normal distribu-
tion with jjL = 1020 and a — 153. We are to find the value of x such that the area under the
normal distribution curve to the right of x is 10%, as shown in Figure 6.49.
Finding x when the area in the
right tail is known.
282 Chapter 6 Continuous Random Variables and the Normal Distribution
First, we find the area under the normal distribution curve to the left of the x value.
Area to the left of the x value = 1.0 - .10 = .9000
To find the z value that corresponds to the required x value, we look for .9000 in the body of
the normal distribution table. The value closest to .9000 in Table IV is .8997, and the corre-
sponding z value is 1.28. Hence, the value of x is computed as
x = fju + za = 1020 + 1.28(153) = 1020 + 195.84 = 1215.84 = 1216
Thus, if Debbie scores 1216 on the SAT, only about 10% of the examinees are expected to
score higher than she does. ■
J EXERCISES
CONCEPTS AND PROCEDURES
6.53 Find the value of z so that the area under the standard normal curve
a. from to z is .4772 and z is positive
b. between and z is (approximately) .4785 and z is negative
c. in the left tail is (approximately) .3565
d. in the right tail is (approximately) .1530
6.54 Find the value of z so that the area under the standard normal curve
a. from to z is (approximately) .1965 and z is positive
b. between and z is (approximately) .2740 and z is negative
c. in the left tail is (approximately) .2050
d. in the right tail is (approximately) .1053
6.55 Determine the value of z so that the area under the standard normal curve
a. in the right tail is .0500 b. in the left tail is .0250
c. in the left tail is .0100 d. in the right tail is .0050
6.56 Determine the value of z so that the area under the standard normal curve
a. in the right tail is .0250 b. in the left tail is .0500
c. in the left tail is .0010 d. in the right tail is .0100
6.57 Let x be a continuous random variable that follows a normal distribution with a mean of 200 and a
standard deviation of 25.
a. Find the value of x so that the area under the normal curve to the left of x is approximately .6330.
b. Find the value of x so that the area under the normal curve to the right of x is approximately .05.
c. Find the value of x so that the area under the normal curve to the right of x is .8051.
d. Find the value of x so that the area under the normal curve to the left of x is .0150.
e. Find the value of x so that the area under the normal curve between jx and x is .4525 and the value
of x is less than /jl.
f. Find the value of x so that the area under the normal curve between /j, and x is approximately
.4800 and the value of x is greater than /jl.
6.58 Let x be a continuous random variable that follows a normal distribution with a mean of 550 and a
standard deviation of 75.
a. Find the value of x so that the area under the normal curve to the left of x is .0250.
b. Find the value of x so that the area under the normal curve to the right of x is .9345.
c. Find the value of x so that the area under the normal curve to the right of x is approximately .0275.
d. Find the value of x so that the area under the normal curve to the left of x is approximately .9600.
e. Find the value of x so that the area under the normal curve between /x and x is approximately
.4700 and the value of x is less than /j,.
f. Find the value of x so that the area under the normal curve between /jl and x is approximately .4100
and the value of x is greater than /jl.
■ APPLICATIONS
6.59 Fast Auto Service provides oil and lube service for cars. It is known that the mean time taken for oil and
lube service at this garage is 15 minutes per car and the standard deviation is 2.4 minutes. The management
wants to promote the business by guaranteeing a maximum waiting time for its customers. If a customer's car
6.7 The Normal Approximation to the Binomial Distribution
283
is not serviced within that period, the customer will receive a 50% discount on the charges. The company
wants to limit this discount to at most 5% of the customers. What should the maximum guaranteed waiting
time be? Assume that the times taken for oil and lube service for all cars have a normal distribution.
6.60 The management of a supermarket wants to adopt a new promotional policy of giving a free gift to
every customer who spends more than a certain amount per visit at this supermarket. The expectation of
the management is that after this promotional policy is advertised, the expenditures for all customers at
this supermarket will be normally distributed with a mean of $95 and a standard deviation of $20. If the
management wants to give free gifts to at most 10% of the customers, what should the amount be above
which a customer would receive a free gift?
6.61 According to the records of an electric company serving the Boston area, the mean electricity con-
sumption during winter for all households is 1650 kilowatt-hours per month. Assume that the monthly
electric consumptions during winter by all households in this area have a normal distribution with a mean
of 1650 kilowatt-hours and a standard deviation of 320 kilowatt-hours. The company sent a notice to Bill
Johnson informing him that about 90% of the households use less electricity per month than he does. What
is Bill Johnson's monthly electricity consumption?
6.62 Rockingham Corporation makes electric shavers. The life (period before which a shaver does not
need a major repair) of Model J795 of an electric shaver manufactured by this corporation has a normal
distribution with a mean of 70 months and a standard deviation of 8 months. The company is to deter-
mine the warranty period for this shaver. Any shaver that needs a major repair during this warranty pe-
riod will be replaced free by the company.
a. What should the warranty period be if the company does not want to replace more than 1% of
the shavers?
b. What should the warranty period be if the company does not want to replace more than 5% of
the shavers?
*6.63 A study has shown that 20% of all college textbooks have a price of $90 or higher. It is known that
the standard deviation of the prices of all college textbooks is $9.50. Suppose the prices of all college text-
books have a normal distribution. What is the mean price of all college textbooks?
*6.64 A machine at Keats Corporation fills 64-ounce detergent jugs. The machine can be adjusted to pour,
on average, any amount of detergent into these jugs. However, the machine does not pour exactly the same
amount of detergent into each jug; it varies from jug to jug. It is known that the net amount of detergent
poured into each jug has a normal distribution with a standard deviation of .35 ounce. The quality control
inspector wants to adjust the machine such that at least 95% of the jugs have more than 64 ounces of de-
tergent. What should the mean amount of detergent poured by this machine into these jugs be?
6.7 The Normal Approximation
to the Binomial Distribution
Recall from Chapter 5 that:
1. The binomial distribution is applied to a discrete random variable.
2. Each repetition, called a trial, of a binomial experiment results in one of two possible out-
comes, either a success or a failure.
3. The probabilities of the two (possible) outcomes remain the same for each repetition of the
experiment.
4. The trials are independent.
The binomial formula, which gives the probability of x successes in n trials, is
P(x) = n C x p*q*-*
The use of the binomial formula becomes very tedious when n is large. In such cases, the
normal distribution can be used to approximate the binomial probability. Note that for a bino-
mial problem, the exact probability is obtained by using the binomial formula. If we apply the
normal distribution to solve a binomial problem, the probability that we obtain is an approxima-
tion to the exact probability. The approximation obtained by using the normal distribution is very
close to the exact probability when n is large and p is very close to .50. However, this does not
284 Chapter 6 Continuous Random Variables and the Normal Distribution
mean that we should not use the normal approximation when p is not close to .50. The reason
the approximation is closer to the exact probability when p is close to .50 is that the binomial
distribution is symmetric when p = .50. The normal distribution is always symmetric. Hence,
the two distributions are very close to each other when n is large and p is close to .50. However,
this does not mean that whenever p = .50, the binomial distribution is the same as the normal
distribution because not every symmetric bell-shaped curve is a normal distribution curve.
Normal Distribution as an Approximation to Binomial Distribution Usually, the normal distribu-
tion is used as an approximation to the binomial distribution when np and nq are both greater
than 5 — that is, when
np > 5 and nq > 5
Table 6.5 gives the binomial probability distribution of x for n = 12 and p = .50. This table
is constructed using Table I of Appendix C. Figure 6.50 shows the histogram and the smoothed
polygon for the probability distribution of Table 6.5. As we can observe, the histogram in
Figure 6.50 is symmetric, and the curve obtained by joining the upper midpoints of the rectan-
gles is approximately bell shaped.
Table 6.5 The Binomial Probability
Distribution for n = 12
and p - .50
X
P(x)
.0002
1
.0029
2
.0161
3
.0537
4
.1208
5
.1934
6
.2256
7
.1934
8
.1208
9
.0537
10
.0161
11
.0029
12
.0002
Figure 6.50 Histogram for the probability distribution
of Table 6.5.
Using the normal
approximation to the binomial:
x equals a specific value.
Examples 6-20 through 6-22 illustrate the application of the normal distribution as an ap-
proximation to the binomial distribution.
■ EXAMPLE 6-20
According to an estimate, 50% of the people in the United States have at least one credit card.
If a random sample of 30 persons is selected, what is the probability that 19 of them will have
at least one credit card?
Solution Let n be the total number of persons in the sample, x be the number of persons in
the sample who have at least one credit card, and p be the probability that a person has at least
one credit card. Then, this is a binomial problem with
30,
.50,
1
1
.50,
x = 19, n - x = 30 - 19 = 11
6.7 The Norma] Approximation to the Binomial Distribution 285
From the binomial formula, the exact probability that 19 persons in a sample of 30 have at
least one credit card is
P(19) = 3oC 19 (.50) l9 (.50) n = .0509
Now let us solve this problem using the normal distribution as an approximation to the
binomial distribution. For this example,
np = 30(.50) = 15 and nq = 30(.50) = 15
Because np and nq are both greater than 5, we can use the normal distribution as an approx-
imation to solve this binomial problem. We perform the following three steps.
Step 1. Compute pu and a for the binomial distribution.
To use the normal distribution, we need to know the mean and standard deviation of the
distribution. Hence, the first step in using the normal approximation to the binomial distribu-
tion is to compute the mean and standard deviation of the binomial distribution. As we know
from Chapter 5, the mean and standard deviation of a binomial distribution are given by np
and \/npq, respectively. Using these formulas, we obtain
l± = np = 30(.50) = 15
o- = Vnpq = V30(.50)(.50) = 2.73861279
Step 2. Convert the discrete random variable into a continuous random variable.
The normal distribution applies to a continuous random variable, whereas the binomial
distribution applies to a discrete random variable. The second step in applying the normal
approximation to the binomial distribution is to convert the discrete random variable to a con-
tinuous random variable by making the correction for continuity.
Definition
Continuity Correction Factor The addition of .5 and/or subtraction of .5 from the value(s) of x
when the normal distribution is used as an approximation to the binomial distribution, where x is
the number of successes in n trials, is called the continuity correction factor.
As shown in Figure 6.51, the probability of 19 successes in 30 trials is given by the area of
the rectangle for x = 19. To make the correction for continuity, we use the interval 18.5 to 19.5
for 19 persons. This interval is actually given by the two boundaries of the rectangle for x = 19,
which are obtained by subtracting .5 from 19 and by adding .5 to 19. Thus, P(x = 19) for the bi-
nomial problem will be approximately equal to P(18.5 < x < 19.5) for the normal distribution.
The area contained by the rectangle for
x = 1 9 is approximated by the area under
the curve between 18.5 and 19.5.
L
Figure 6.51
t
11= 15
19
LtL
H= 15
1 1
18.5— 1 !— 1
Step 3. Compute the required probability using the normal distribution.
9.5
As shown in Figure 6.52, the area under the normal distribution curve between x = 18.5
and x = 19.5 will give us the (approximate) probability that 19 persons have at least one credit
286 Chapter 6 Continuous Random Variables and the Normal Distribution
card. We calculate this probability as follows:
Figure 6.52 Area between x = 18.5 and x = 19.5.
The required probability is given by the area under the standard normal curve between z = 1 .28
and z = 1-64. This area is obtained by subtracting the area to the left of z = 1-28 from the
area to the left of z = 1 .64. From Table IV of Appendix C, the area to the left of z = 1 -28 is
.8997 and the area to the left of z = 1-64 is .9495. Hence, the required probability is
P(18.5 < x < 19.5) = P(1.28 < z < 1.64) = .9495 - .8997 = .0498
Thus, based on the normal approximation, the probability that 19 persons in a sample of
30 will have at least one credit card is approximately .0498. Earlier, using the binomial for-
mula, we obtained the exact probability .0509. The error due to using the normal approxima-
tion is .0509 - .0498 = .0011. Thus, the exact probability is underestimated by .0011 if the
normal approximation is used. I
Remember ► When applying the normal distribution as an approximation to the binomial distribution, always
make a correction for continuity. The continuity correction is made by subtracting .5 from the
lower limit of the interval and/or by adding .5 to the upper limit of the interval. For example,
the binomial probability P(7 < x < 12) will be approximated by the probability P(6.5 ^ x <
12.5) for the normal distribution; the binomial probability P(x & 9) will be approximated by
the probability P(x > 8.5) for the normal distribution; and the binomial probability P(x < 10)
will be approximated by the probability P(x < 10.5) for the normal distribution. Note that the
probability P(x > 9) has only the lower limit of 9 and no upper limit, and the probability
P(x ^10) has only the upper limit of 10 and no lower limit.
■ EXAMPLE 6-21
According to a joint reader survey by USATODAY.com and TripAdvisor.com, 34% of the peo-
ple surveyed said that the first thing they do after checking into a hotel is to adjust the ther-
mostat (USA TODAY, June 12, 2009). Suppose that this result is true for the current popula-
tion of all adult Americans who stay in hotels. What is the probability that in a random sample
of 400 adult Americans who stay in hotels, 115 to 130 will say that the first thing they do af-
ter checking into a hotel is to adjust the thermostat?
Solution Let n be the total number of adults in the sample, x be the number of adults in the
sample who say that the first thing they do after checking into a hotel is to adjust the thermo-
stat, and p be the probability that an adult says that the first thing he or she does after check-
ing into a hotel is to adjust the thermostat. Then, this is a binomial problem with
Using the normal approxima-
tion to the binomial: x assumes
a value in an interval
n = 400, p = .34, and q = 1 - .34 = .66
6.7 The Norma] Approximation to the Binomial Distribution 287
We are to find the probability of 115 to 130 successes in 400 trials. Because n is large, it is
easier to apply the normal approximation than to use the binomial formula. We can check that
np and nq are both greater than 5. The mean and standard deviation of the binomial distribu-
tion are, respectively,
(ji = np = 400 (.34) = 136
a = Vnpq = V400(.34)(.66) = 9.47417543
To make the continuity correction, we subtract .5 from 115 and add .5 to 130 to obtain the inter-
val 114.5 to 130.5. Thus, the probability that 115 to 130 out of 400 adults will say that the first
thing they do after checking into a hotel is to adjust the thermostat is approximated by the area un-
der the normal distribution curve from x = 1 14.5 to x = 130.5. This area is shown in Figure 6.53.
Forx = 114.5:
Forx = 130.5:
114.5 - 136
9.47417543
130.5 - 136
9.47417543
-2.27
.58
Shaded area ■
is .2694
Figure 6.53 Area between x = 114.5 and x = 130.5.
114.5
130.5 136
i — r
.58
-2.27
The required probability is given by the area under the standard normal curve between
z = —2.27 and z = —.58. This area is obtained by taking the difference between the areas
under the standard normal curve to the left of z = —2.27 and to the left of z = —.58. From
Table IV in Appendix C, the area to the left of z = —2.27 is .0116, and the area to the left of
z = —.58 is .2810. Hence, the required probability is
P(114.5 < x < 130.5) = P(-2.27 < z < -.58) = .2810 - .0116 = .2694
Thus, the probability that 1 15 to 130 adults in a sample of 400 will say that the first thing they
do after checking into a hotel is to adjust the thermostat is approximately .2694. H
EXAMPLE 6-22
According to an American Laser Centers survey, 32% of adult men said that their stomach is
the least favorite part of their body (USA TODAY, March 10, 2009). Assume that this percent-
age is true for the current population of adult men. What is the probability that 170 or more adult
men in a random sample of 500 will say that their stomach is the least favorite part of their body?
Solution Let n be the sample size, x be the number of adult men in the sample who will
say that their stomach is the least favorite part of their body, and p be the probability that a
randomly selected adult man holds this opinion. Then, this is a binomial problem with
Using the normal
approximation to the binomial:
x is greater than or equal
to a value.
n = 500, p = .32, and q = 1 - .32 = .68
We are to find the probability of 170 or more successes in 500 trials. The mean and standard
deviation of the binomial distribution are, respectively,
lL = np = 500 (.32) = 160
o- = Vnpq = V500(.32)(.68)
10.43072385
288
Chapter 6 Continuous Random Variables and the Normal Distribution
For the continuity correction, we subtract .5 from 170, which gives 169.5. Thus, the probabil-
ity that 170 or more of the adult men in a random sample of 500 will hold the said opinion is
approximated by the area under the normal distribution curve to the right of x = 169.5, as
shown in Figure 6.54.
169.5 - 160
Forx= 169.5: z = =.91
10.43072385
To find the required probability, we find the area to the left of z =.91 and subtract this
area from 1.0. From Table IV in Appendix C, the area to the left of z =.91 is .8186.
Hence,
P(x > 169.5) = P(z > .91) = 1.0 - .8186 = .1814
Thus, the probability that 170 or more American adults in a random sample of 500 will say
that their stomach is the least favorite part of their body is approximately .1814. I
EXERCISES
CONCEPTS AND PROCEDURES
6.65 Under what conditions is the normal distribution usually used as an approximation to the binomial
distribution?
6.66 For a binomial probability distribution, n = 20 and p = .60.
a. Find the probability P(x = 14) by using the table of binomial probabilities (Table I of
Appendix C).
b. Find the probability P(x = 14) by using the normal distribution as an approximation to the bino-
mial distribution. What is the difference between this approximation and the exact probability
calculated in part a?
6.67 For a binomial probability distribution, n = 25 and p = .40.
a. Find the probability P(8 £ x S 13) by using the table of binomial probabilities (Table I of
Appendix C).
b. Find the probability P(8 < x < 13) by using the normal distribution as an approximation to the
binomial distribution. What is the difference between this approximation and the exact probability
calculated in part a?
6.68 For a binomial probability distribution, n = 80 and p = .50. Let x be the number of successes in
80 trials.
a. Find the mean and standard deviation of this binomial distribution.
b. Find P(x a 42) using the normal approximation.
c. Find P(41 £ x < 48) using the normal approximation.
6.69 For a binomial probability distribution, n = 120 and p = .60. Let x be the number of successes in
120 trials.
a. Find the mean and standard deviation of this binomial distribution.
b. Find P(x < 69) using the normal approximation.
c. Find P(67 £ x £ 73) using the normal approximation.
6.7 The Normal Approximation to the Binomial Distribution 289
6.70 Find the following binomial probabilities using the normal approximation.
a. n = 140, p = .45, P(x = 67)
b. n = 100, p = .55, P(52 < x < 60)
c. n = 90, p = .42, P(x > 40)
d. n = 104, p = .75, P(x < 72)
6.71 Find the following binomial probabilities using the normal approximation.
a. n = 70, p = .30, P(x = 18)
b. n = 200, p = .70, P(133 < x < 145)
c n = 85, p = .40, P(x > 30)
d. n = 150, p = .38, P(x < 62)
■ APPLICATIONS
6.72 According to a May 27, 2009 Minneapolis Star- Tribune article (Source: http://www.startribune.com/
politics/45797562.html), 78% of U.S. households have at least one credit card. Find the probability that
in a random sample of 500 U.S. households, 375 to 385 households have at least one credit card.
6.73 A 2007 article states that 4.8% of U.S. households are "linguistically isolated," which means that
all members of the household aged 14 years and older have difficulty speaking English (Source:
http://www.antara.co.id/en/arc/2007/9/12/five-percent-of-us-families-dont-speak-english-report/).
Assume that this percentage is true for the current population of U.S. households. Find the probability
that in a random sample of 750 U.S. households, more than 45 would be classified as "linguistically
isolated."
6.74 According to the 2008 ARIS American Religious Survey (Source: www.americanreligionsurvey-
aris.org), 3.52% of U.S. adults classify themselves as "non-denominational Christians." Assume that this
percentage is true for the current population of U.S. adults. Find the probability that in a random sample
of 600 U.S. adults, the number who classify themselves as non-denominational Christians is
a. exactly 25 b. 13 to 19 c. at least 27
6.75 In the Energy Information Administration report The Effect of Income on Appliances in U.S. House-
holds (Source: http://www.eia.doe.gov/emeu/recs/appliances/appliances.html), it is noted that 29% of hous-
ing units with an annual income in the $15,000 to $29,999 range own a large-screen television. Assum-
ing that this is true for the current population of housing units with an annual income in the $15,000 to
$29,999 range, find the probability that in a random sample of 400 such housing units, the number that
have a large screen television is
a. exactly 110 b. 124 to 135 c. no more than 105
6.76 During the 2009 edition of the reality show Britain 's Got Talent, runner-up and Internet singing sen-
sation Susan Boyle obtained 20.2% of the first-place votes. Suppose that this percentage would hold true
for all potential voters (note: the population of interest would be all viewers of ITV, which carries the
show). Find the probability that, in a random sample of 250 potential voters, the number who would vote
for Susan Boyle is
a. exactly 57 b. 35 to 41 c. at least 60
6.77 An office supply company conducted a survey before marketing a new paper shredder designed
for home use. In the survey, 80% of the people who tried the shredder were satisfied with it. Because
of this high satisfaction rate, the company decided to market the new shredder. Assume that 80% of
all people are satisfied with this shredder. During a certain month, 100 customers bought this shred-
der. Find the probability that of these 100 customers, the number who are satisfied is
a. exactly 75 b. 73 or fewer c. 74 to 85
6.78 Johnson Electronics makes calculators. Consumer satisfaction is one of the top priorities of the com-
pany's management. The company guarantees the refund of money or a replacement for any calculator
that malfunctions within two years from the date of purchase. It is known from past data that despite all
efforts, 5% of the calculators manufactured by this company malfunction within a 2-year period. The com-
pany recently mailed 500 such calculators to its customers.
a. Find the probability that exactly 29 of the 500 calculators will be returned for refund or replace-
ment within a 2-year period.
b. What is the probability that 27 or more of the 500 calculators will be returned for refund or
replacement within a 2-year period?
c. What is the probability that 15 to 22 of the 500 calculators will be returned for refund or
replacement within a 2-year period?
290 Chapter 6 Continuous Random Variables and the Normal Distribution
6.79 Hurbert Corporation makes font cartridges for laser printers that it sells to Alpha Electronics Inc.
The cartridges are shipped to Alpha Electronics in large volumes. The quality control department at
Alpha Electronics randomly selects 100 cartridges from each shipment and inspects them for being good
or defective. If this sample contains 7 or more defective cartridges, the entire shipment is rejected. Hurbert
Corporation promises that of all the cartridges, only 5% are defective.
a. Find the probability that a given shipment of cartridges received by Alpha Electronics will be
accepted.
b. Find the probability that a given shipment of cartridges received by Alpha Electronics will not be
accepted.
USES AND MISUSES... DON'T LOSE YOUR MEMORY
As discussed in the previous chapter, the Poisson distribution gives
the probability of a specified number of events occurring in a time
interval. The Poisson distribution provides a model for the number of
emails a server might receive during a certain time period or the num-
ber of people arriving in line at a bank during lunch hour. These are
nice to know for planning purposes, but sometimes we want to know
the specific times at which emails or customers arrive. These times
are governed by a special continuous probability distribution with cer-
tain unusual properties. This distribution is called the exponential dis-
tribution, and it is derived from the Poisson probability distribution.
Suppose you are a teller at a bank, and a customer has just
arrived. You know that the customers arrive according to a Poisson
process with a rate of A customers per hour. Your boss might care
how many customers arrive on average during a given time interval
to ensure there are enough tellers available to handle the customers
efficiently; you are more concerned with the time when the next cus-
tomer will arrive. Remember that the probability that x customers
arrive in an interval of length r is
(Af)*e- Af
P(x)
x\
The probability that a customer arrives within time f is 1 minus the
probability that no customer arrives within time f. Hence,
P(customer arrives within time r) = 1 - P(0)
(Af)°e- A '
If the bank receives an average of 15 customers per hour— an
average of one every 4 minutes— and a customer has just ar-
rived, the probability that a customer arrives within 4 minutes is
1 - e _Af = 1 - e _(15/60)4 = .6 3 21. In the same way, the prob-
ability that a customer arrives within 8 minutes is .8647.
Let us say that a customer arrived and went to your co-worker's
window. No additional customer arrived within the next 2 minutes—
an event with probability .6065— and you dozed off for 2 more min-
utes. When you open your eyes, you see that a customer has not ar-
rived yet. What is the probability that a customer arrives within the
next 4 minutes? From the calculation above, you might say that the
answer is .8647. After all, you know that a customer arrived 8 min-
utes earlier. But .8647 is not the correct answer.
The exponential distribution, which governs the time between ar-
rivals of a Poisson process, has a property called the memoryless prop-
erty. For you as a bank teller, this means that if you know a customer
has not arrived during the past 4 minutes, then the clock is reset to
zero, as if the previous customer had just arrived. So even after your
nap, the probability that a customer arrives within 4 minutes is .6321.
This interesting property reminds us again that we should be careful
when we use mathematics to model real-world phenomena.
1
0!
1
Glossary
Continuity correction factor Addition of .5 and/or subtraction of
.5 from the value(s) of x when the normal distribution is used as an
approximation to the binomial distribution, where x is the number
of successes in n trials.
Continuous random variable A random variable that can assume
any value in one or more intervals.
Normal probability distribution The probability distribution of
a continuous random variable that, when plotted, gives a specific
bell-shaped curve. The parameters of the normal distribution are the
mean /j, and the standard deviation a.
Standard normal distribution The normal distribution with yu, =
and cr=l. The units of the standard normal distribution are denoted
by z.
Z value or z score The units of the standard normal distribution
that are denoted by z.
Supplementary Exercises 291
Supplementary Exercises
6.80 The management at Ohio National Bank does not want its customers to wait in line for service for
too long. The manager of a branch of this bank estimated that the customers currently have to wait an av-
erage of 8 minutes for service. Assume that the waiting times for all customers at this branch have a nor-
mal distribution with a mean of 8 minutes and a standard deviation of 2 minutes.
a. Find the probability that a randomly selected customer will have to wait for less than 3
minutes.
b. What percentage of the customers have to wait for 10 to 13 minutes?
c. What percentage of the customers have to wait for 6 to 12 minutes?
d. Is it possible that a customer may have to wait longer than 16 minutes for service? Explain.
6.81 A company that has a large number of supermarket grocery stores claims that customers who pay by
personal check spend an average of $87 on groceries at these stores with a standard deviation of $22. Assume
that the expenses incurred on groceries by all such customers at these stores are normally distributed.
a. Find the probability that a randomly selected customer who pays by check spends more than
$1 14 on groceries.
b. What percentage of customers paying by check spend between $40 and $60 on groceries?
c. What percentage of customers paying by check spend between $70 and $105?
d. Is it possible for a customer paying by check to spend more than $185? Explain.
6.82 At Jen and Perry Ice Cream Company, the machine that fills 1 -pound cartons of Top Flavor ice cream
is set to dispense 16 ounces of ice cream into every carton. However, some cartons contain slightly less
than and some contain slightly more than 16 ounces of ice cream. The amounts of ice cream in all such
cartons have a normal distribution with a mean of 16 ounces and a standard deviation of .18 ounce.
a. Find the probability that a randomly selected carton contains 16.20 to 16.50 ounces of ice cream.
b. What percentage of such cartons contain less than 15.70 ounces of ice cream?
c. Is it possible for a carton to contain less than 15.20 ounces of ice cream? Explain.
6.83 A machine at Kasem Steel Corporation makes iron rods that are supposed to be 50 inches long. How-
ever, the machine does not make all rods of exactly the same length. It is known that the probability dis-
tribution of the lengths of rods made on this machine is normal with a mean of 50 inches and a standard
deviation of .06 inch. The rods that are either shorter than 49.85 inches or longer than 50.15 inches are
discarded. What percentage of the rods made on this machine are discarded?
6.84 Jenn Bard, who lives in San Francisco Bay area, commutes by car from home to work. She has found
out that it takes her an average of 28 minutes for this commute in the morning. However, due to the vari-
ability in the traffic situation every morning, the standard deviation of these commutes is 5 minutes. Sup-
pose the population of her morning commute times has a normal distribution with a mean of 28 minutes
and a standard deviation of 5 minutes. Jenn has to be at work by 8:30 a.m. every morning. By what time
must she leave home in the morning so that she is late for work at most 1% of the time?
6.85 The print on the package of 100- watt General Electric soft- white lightbulbs states that these bulbs
have an average life of 750 hours. Assume that the lives of all such bulbs have a normal distribution with
a mean of 750 hours and a standard deviation of 50 hours.
a. Let x be the life of such a lightbulb. Find x so that only 2.5% of such lightbulbs have lives
longer than this value.
b. Let x be the life of such a lightbulb. Find x so that about 80% of such lightbulbs have lives
shorter than this value.
6.86 Major League Baseball rules require that the balls used in baseball games must have circumferences
between 9 and 9.25 inches. Suppose the balls produced by the factory that supplies balls to Major League
Baseball have circumferences normally distributed with a mean of 9.125 inches and a standard deviation
of .06 inch. What percentage of these baseballs fail to meet the circumference requirement?
6.87 Mong Corporation makes auto batteries. The company claims that 80% of its LL70 batteries are good
for 70 months or longer.
a. What is the probability that in a sample of 100 such batteries, exactly 85 will be good for
70 months or longer?
b. Find the probability that in a sample of 100 such batteries, at most 74 will be good for
70 months or longer.
c. What is the probability that in a sample of 100 such batteries, 75 to 87 will be good for
70 months or longer?
d. Find the probability that in a sample of 100 such batteries, 72 to 77 will be good for 70 months
or longer.
292 Chapter 6 Continuous Random Variables and the Normal Distribution
6.88 Stress on the job is a major concern of a large number of people who go into managerial positions.
It is estimated that 80% of the managers of all companies suffer from job-related stress.
a. What is the probability that in a sample of 200 managers of companies, exactly 150 suffer from
job-related stress?
b. Find the probability that in a sample of 200 managers of companies, at least 170 suffer from
job-related stress.
c. What is the probability that in a sample of 200 managers of companies, 165 or fewer suffer
from job-related stress?
d. Find the probability that in a sample of 200 managers of companies, 164 to 172 suffer from
job-related stress.
■ Advanced Exercises
6.89 It is known that 15% of all homeowners pay a monthly mortgage of more than $2500 and that the
standard deviation of the monthly mortgage payments of all homeowners is $350. Suppose that the monthly
mortgage payments of all homeowners have a normal distribution. What is the mean monthly mortgage
paid by all homeowners?
6.90 At Jen and Perry Ice Cream Company, a machine fills 1 -pound cartons of Top Flavor ice cream. The
machine can be set to dispense, on average, any amount of ice cream into these cartons. However, the ma-
chine does not put exactly the same amount of ice cream into each carton; it varies from carton to carton.
It is known that the amount of ice cream put into each such carton has a normal distribution with a stan-
dard deviation of .18 ounce. The quality control inspector wants to set the machine such that at least 90%
of the cartons have more than 16 ounces of ice cream. What should be the mean amount of ice cream put
into these cartons by this machine?
6.91 Two companies, A and B, drill wells in a rural area. Company A charges a flat fee of $3500 to drill
a well regardless of its depth. Company B charges $1000 plus $12 per foot to drill a well. The depths of
wells drilled in this area have a normal distribution with a mean of 250 feet and a standard deviation of
40 feet.
a. What is the probability that Company B would charge more than Company A to drill a well?
b. Find the mean amount charged by Company B to drill a well.
6.92 Otto is trying out for the javelin throw to compete in the Olympics. The lengths of his javelin throws
are normally distributed with a mean of 290 feet and a standard deviation of 10 feet. What is the proba-
bility that the longest of three of his throws is 320 feet or more?
6.93 Lori just bought a new set of four tires for her car. The life of each tire is normally distributed with a
mean of 45,000 miles and a standard deviation of 2000 miles. Find the probability that all four tires will last
for at least 46,000 miles. Assume that the life of each of these tires is independent of the lives of other tires.
6.94 The Jen and Perry Ice Cream company makes a gourmet ice cream. Although the law allows ice
cream to contain up to 50% air, this product is designed to contain only 20% air. Because of variability
inherent in the manufacturing process, management is satisfied if each pint contains between 18% and
22% air. Currently two of Jen and Perry's plants are making gourmet ice cream. At Plant A, the mean
amount of air per pint is 20% with a standard deviation of 2%. At Plant B, the mean amount of air per
pint is 19% with a standard deviation of 1%. Assuming the amount of air is normally distributed at both
plants, which plant is producing the greater proportion of pints that contain between 18% and 22% air?
6.95 The highway police in a certain state are using aerial surveillance to control speeding on a highway
with a posted speed limit of 55 miles per hour. Police officers watch cars from helicopters above a straight
segment of this highway that has large marks painted on the pavement at 1-mile intervals. After the po-
lice officers observe how long a car takes to cover the mile, a computer estimates that car's speed. As-
sume that the errors of these estimates are normally distributed with a mean of and a standard deviation
of 2 miles per hour.
a. The state police chief has directed his officers not to issue a speeding citation unless the aerial
unit's estimate of speed is at least 65 miles per hour. What is the probability that a car traveling
at 60 miles per hour or slower will be cited for speeding?
b. Suppose the chief does not want his officers to cite a car for speeding unless they are 99% sure
that it is traveling at 60 miles per hour or faster. What is the minimum estimate of speed at
which a car should be cited for speeding?
6.96 Ashley knows that the time it takes her to commute to work is approximately normally distributed
with a mean of 45 minutes and a standard deviation of 3 minutes. What time must she leave home in the
morning so that she is 95% sure of arriving at work by 9 a.m.?
Supplementary Exercises 293
6.97 A soft-drink vending machine is supposed to pour 8 ounces of the drink into a paper cup. However,
the actual amount poured into a cup varies. The amount poured into a cup follows a normal distribution
with a mean that can be set to any desired amount by adjusting the machine. The standard deviation of
the amount poured is always .07 ounce regardless of the mean amount. If the owner of the machine wants
to be 99% sure that the amount in each cup is 8 ounces or more, to what level should she set the mean?
6.98 A newspaper article reported that the mean mathematics score on SAT for students from a local high
school was 500 and that 20% of the students scored below 430. Assume that the SAT scores for students
from this school follow a normal distribution.
a. Find the standard deviation of the mathematics SAT scores for students from this school.
b. Find the percentage of students at this school whose mathematics SAT scores were above 520.
6.99 Alpha Corporation is considering two suppliers to secure the large amounts of steel rods that it uses.
Company A produces rods with a mean diameter of 8 mm and a standard deviation of .15 mm and sells
10,000 rods for $400. Company B produces rods with a mean diameter of 8 mm and a standard deviation
of .12 mm and sells 10,000 rods for $460. A rod is usable only if its diameter is between 7.8 mm and
8.2 mm. Assume that the diameters of the rods produced by each company have a normal distribution.
Which of the two companies should Alpha Corporation use as a supplier? Justify your answer with ap-
propriate calculations.
6.100 A gambler is planning to make a sequence of bets on a roulette wheel. Note that a roulette wheel
has 38 numbers, of which 18 are red, 18 are black, and 2 are green. Each time the wheel is spun, each of
the 38 numbers is equally likely to occur. The gambler will choose one of the following two sequences.
Single-number bet: The gambler will bet $5 on a particular number before each spin. He will win a net
amount of $175 if that number comes up and lose $5 otherwise.
Color bet: The gambler will bet $5 on the red color before each spin. He will win a net amount of $5 if
a red number comes up and lose $5 otherwise.
a. If the gambler makes a sequence of 25 bets, which of the two betting schemes offers him a
better chance of coming out ahead (winning more money than losing) after the 25 bets?
b. Now compute the probability of coming out ahead after 25 single-number bets of $5
each and after 25 color bets of $5 each. Do these results confirm your guess in part a?
(Before using an approximation to find either probability, be sure to check whether it is
appropriate.)
6.101 A charter bus company is advertising a singles outing on a bus that holds 60 passengers. The com-
pany has found that, on average, 10% of ticket holders do not show up for such trips; hence, the company
routinely overbooks such trips. Assume that passengers act independently of one another.
a. If the company sells 65 tickets, what is the probability that the bus can hold all the ticket holders
who actually show up? In other words, find the probability that 60 or fewer passengers show up.
b. What is the largest number of tickets the company can sell and still be at least 95% sure that
the bus can hold all the ticket holders who actually show up?
6.102 The amount of time taken by a bank teller to serve a randomly selected customer has a normal dis-
tribution with a mean of 2 minutes and a standard deviation of .5 minute.
a. What is the probability that both of two randomly selected customers will take less than
1 minute each to be served?
b. What is the probability that at least one of four randomly selected customers will need more
than 2.25 minutes to be served?
6.103 Suppose you are conducting a binomial experiment that has 15 trials and the probability of success
of .02. According to the sample size requirements, you cannot use the normal distribution to approximate
the binomial distribution in this situation. Use the mean and standard deviation of this binomial distribu-
tion and the empirical rule to explain why there is a problem in this situation. (Note: Drawing the graph
and marking the values that correspond to the empirical rule is a good way to start.)
6.104 A variation of a roulette wheel has slots that are not of equal size. Instead, the width of any slot is
proportional to the probability that a standard normal random variable z takes on a value between a and
(a + .1), where a = —3.0, —2.9,-2.8, . . . , 2.9, 3.0. In other words, there are slots for the intervals
(-3.0, -2.9), (-2.9, -2.8), (-2.8, -2.7) through (2.9, 3.0). There is one more slot that represents the
probability that z falls outside the interval ( — 3.0, 3.0). Find the following probabilities.
a. The ball lands in the slot representing (.3, .4).
b. The ball lands in any of the slots representing ( — .1, .4).
c. In at least one out of five games, the ball lands in the slot representing ( — .1, .4).
d. In at least 100 out of 500 games, the ball lands in the slot representing (.4, .5).
294
Chapter 6 Continuous Random Variables and the Normal Distribution
6.105 Refer to Exercise 6.97. In that exercise, suppose the mean is set to be 8 ounces, but the standard
deviation is unknown. The cups used in the machine can hold up to 8.2 ounces, but these cups will over-
flow if more than 8.2 ounces is dispensed by the machine. What is the smallest possible standard devia-
tion that will result in overflows occurring 3% of the time?
Self-Review Test
1. The normal probability distribution is applied to
a. a continuous random variable b. a discrete random variable c. any random variable
2. For a continuous random variable, the probability of a single value of x is always
a. zero b. 1.0 c. between and 1
3. Which of the following is not a characteristic of the normal distribution?
a. The total area under the curve is 1.0.
b. The curve is symmetric about the mean.
c. The two tails of the curve extend indefinitely.
d. The value of the mean is always greater than the value of the standard deviation.
4. The parameters of a normal distribution are
a. /x, z, and cr b. /x and cr c. p., x, and cr
5. For the standard normal distribution,
a. /x — and cr = 1 b. fi = 1 and cr = c. p = 100 and cr = 10
6. The z value for p for a normal distribution curve is always
a. positive b. negative c.
7. For a normal distribution curve, the z value for an x value that is less than p is always
a. positive b. negative c.
8. Usually the normal distribution is used as an approximation to the binomial distribution when
a. n > 30 b. np > 5 and nq > 5 c. n > 20 and p = .50
9. Find the following probabilities for the standard normal distribution.
a. P(.85 < z < 2.33) b. P(- 2.97 < z £ 1.49) c. P(z < - 1 .29) d. P(z > -.74)
10. Find the value of z for the standard normal curve such that the area
a. in the left tail is .1000 b. between and z is .2291 and z is positive
c. in the right tail is .0500 d. between and z is .3571 and z is negative
11. In a National Highway Transportation Safety Administration (NHSTA) report, data provided to the
NHSTA by Goodyear stated that the average tread life of properly inflated automobile tires is 45,000 miles
(Source: http://www.nhtsa.dot.gOv/cars/rules/rulings/TPMS_FMVSS_Nol38/part5.5.html). Suppose that
the current distribution of tread life of properly inflated automobile tires is normally distributed with a
mean of 45,000 miles and a standard deviation of 2360 miles.
a. Find the probability that a randomly selected automobile tire has a tread life between 42,000
and 46,000 miles.
b. What is the probability that a randomly selected automobile tire has a tread life of less than
38,000 miles?
c. What is the probability that a randomly selected automobile tire has a tread life of more than
50,000 miles?
d. Find the probability that a randomly selected automobile tire has a tread life between 46,500
and 47,500 miles.
12. Refer to Problem 11.
a. Suppose that 6% of all automobile tires with the longest tread life have a tread life of at least x
miles. Find the value of x.
b. Suppose that 2% of all automobile tires with the shortest tread life have a tread life of at most x
miles. Find the value of x.
13. Gluten sensitivity, which is also known as wheat intolerance, affects approximately 15% of people. The
condition involves great difficulty in digesting wheat, but is not the same as wheat allergy, which has much more
severe reactions (Source: http://www.foodintol.com/wheat.asp). A random sample of 800 individuals is selected.
a. Find the probability that the number of individuals in this sample who have wheat intolerance is
i. exactly 115 ii. 103 to 142 iii. at least 107
iv. at most 100 v. between 111 and 123
Mini-Projects 295
b. Find the probability that at least 675 of the individuals in this sample do not have wheat
intolerance.
c. Find the probability that 682 to 697 of the individuals in this sample do not have wheat
intolerance.
Mini-Projects
■ MINI-PROJECT 6-1
Consider the data on heights of NBA players that accompany this text (see Appendix B).
a. Use statistical software to obtain a histogram. Do these heights appear to be symmetrically
distributed? If not, in which direction do they seem to be skewed?
b. Compute /x and a for heights of all players.
c. What percentage of these heights lie in the interval ix — a to /x + cr? What about in the interval
/x — 2a to /x + 2<x? In the interval /x — 3a to /x + 3cr?
d. How do the percentages in part c compare to the corresponding percentages for a normal distri-
bution (68.26%, 95.44%, and 99.74%, respectively)?
e. Use statistical software to select three random samples of 20 players each. Create a histogram and
a dotplot of heights for each sample, and calculate the mean and standard deviation of heights for
each sample. How well do your graphs and summary statistics match up with the corresponding
population graphs and parameter values obtained in earlier parts? Does it seem reasonable that
they might not match up very well?
■ MINI-PROJECT 6-2
Consider the data on weights of NBA players (see Appendix B).
a. Use statistical software to obtain a histogram. Do these weights appear to be symmetrically
distributed? If not, in which direction do they seem to be skewed?
b. Compute /j, and a for weights of all players.
c. What percentage of these weights lie in the interval /x — cr to /x + cr? What about in the interval
/x — 2a to /x + 2cr? In the interval jx — 3a to ix + 3cr?
d. How do the percentages in part c compare to the corresponding percentages for a normal distri-
bution (68.26%, 95.44%, and 99.74%, respectively)?
e. Use statistical software to select three random samples of 20 players each. Create a histogram and
a dotplot of weights for each sample, and calculate the mean and standard deviation of weights
for each sample. How well do your graphs and summary statistics match up with the correspon-
ding population graphs and parameter values obtained in earlier parts? Does it seem reasonable
that they might not match up very well?
■ MINI-PROJECT 6-3
The National Oceanic and Atmospheric Administration (NOAA) web site has daily historical data precip-
itation amounts, as well as the minimum and maximum temperatures available for a large number of
weather stations throughout the United States. For the purpose of this Mini-Project, you will need to down-
load 2 consecutive months of data, 1 month at a time. To obtain the data, go to http://www7.ncdc.noaa.gov/
IPS/coop/coop.html and choose your location and month of interest. Answer the following questions with
regard to the maximum daily temperature.
a. Use statistical software to obtain a histogram and a dotplot for your data. Comment on the shape
of the distribution as observed from these graphs.
b. Calculate x and .v.
c. What percentage of the temperatures are in the interval x — s to x + s?
d. What percentage are in the interval x — 2s to x + 2i?
e. How do these percentages compare to the corresponding percentages for a normal distribution
(68.26% and 95.44%, respectively)?
f. Now find the minimum temperatures in your town for 60 days by using the same source that you
used to find the maximum temperatures or by using a different source. Then repeat parts a through
e for this data set.
296 Chapter 6 Continuous Random Variables and the Normal Distribution
DECIDE FOR YOURSELF
Deciding About the Shape
of a Distribution
Reporting summary measures such as the mean, median, and stan-
dard deviation has become very common in modem life. Many com-
panies, government agencies, and so forth will report the mean and
standard deviation of a variable, but they will very rarely provide
information on the shape of the distribution of that variable. In
Chapters 5 and 6, you have learned some basic properties of some
distributions that can help you to decide if a specific type of distri-
bution is a good fit for a set of data.
According to the National Diet and Nutrition Survey: Adults
Aged 19 to 64, British men spend an average of 2.15 hours per day
in moderate- or high-intensity physical activity. The standard devia-
tion of these activity times for this sample was 3.59 hours. {Source:
http://www.food.gov.uk/multimedia/pdfs/ndnsfour.pdf.) Can we infer
that these activity times could follow a normal distribution? The fol-
lowing questions may provide an answer.
1. Sketch a normal curve marking the points representing 1, 2, and
3 standard deviations above and below the mean, and calculate the
values at these points using a mean of 2.15 hours and a standard
deviation of 3.59 hours.
2. Examine the curve with your calculations. Explain why it is
impossible for this distribution to be normal based on your graph and
calculations.
3. Considering the variable being measured, is it more likely that
the distribution is skewed to the left or that it is skewed to the right?
Explain why.
4. Suppose that the standard deviation for this sample was .70 hour
instead of 3.59 hours, which makes it numerically possible for the
distribution to be normal. Again, considering the variable being
measured, explain why the normal distribution is still not a logical
choice for this distribution.
ECHNOLOGY
INSTRUCTION
Normal and Inverse Normal Probabilities
norma lcdfC -e99, 1
25, 100, 15>
. 9522096696
Screen 6.1
1. For a given mean /x and standard deviation cr, to find the probability that a normal random
variable x lies below b, select DISTR >normalcdf(-E99, b, fi, cr) and press ENTER.
(See Screen 6.1.)
2. For a given mean fi and standard deviation cr, to find the probability that a normal random
variable x lies above a, select DISTR >normalcdf(a, E99, fi, a) and press ENTER.
3. For a given mean /a and standard deviation cr, to find the probability that a normal random
variable x lies between a and b, select DISTR >normalcdf(a, b, fi, cr) and press ENTER.
4. To find a value of a for a normal random variable x with mean j± and standard deviation cr
such that the probability of x being less than a is p, select DISTR >invNorm(p, fi, cr) and
press ENTER.
Note: To type E99, press 2nd >comma key (which is the key just above the 7 key). The
function is labeled EE, but only E is displayed on the screen. Then type 9 twice. For -E99,
press (-) key (which is to the right of the decimal key) before E99.
1. For a given mean fi and standard deviation cr, to find the probability that a normal random
variable x lies below a, select Calc >Probability Distributions >Normal. Select Cumulative
probability, and enter the mean /jl and the standard deviation cr. Select Input constant and
enter a, then select OK. (See Screens 6.2 and 6.3.)
Technology Instruction 297
2. To find a value of a for a normal random variable x with mean /j, and standard deviation
a such that the probability of x being less than a is p, select Calc >Probability Distrib-
utions >Normal. Select Inverse cumulative probability and enter the mean /jl and the
standard deviation cr. Select Input constant and enter a, then select OK.
Normal Distribution
<~ Probability density
• Cumulative probability
<"* Inverse cumulative probability
Mean: [100
Standard deviation: fls
r Input column:
Optional storage:
f* Input constant: |l25
Optional storage:
OK | Cancel
Screen 6.2
Select
Help
Cumulative Distribution Function
Normal with mean = 100 and standard deviation = 15
x P[ X <= x )
125 0.952210
Screen 6.3
1. For a given mean p, and standard deviation cr, to find the probability that a normal random
variable x lies below b, type =NORMDIST(Z>, fi, a, 1). (See Screen 6.4.)
2. For a given mean p and standard deviation cr, to find the probability that a normal random
variable x lies above a, type =1— NORMDIST(a, fi, cr, 1).
3. For a given mean /jl and standard deviation cr, to find the probability that a normal random
variable x lies between a and b, type =NORMDIST(2>, fi, a, l)-NORMDIST(a, fi, a, 1).
4. To find a value of a for a normal random variable x with mean jx and standard deviation cr
such that the probability of x being less than a is p, type =NORMINV(/?, fi, a).
A
B | C
D
1
Mean
100
2
Std. Dev.
3
4
P(X<125)
=NORMDIST(125,100,15,1)
Screen 6.4
298
Chapter 6 Continuous Random Variables and the Normal Distribution
TECHNOLOGY ASSIGNMENTS
Note: Virtually all of the problems in this chapter can be completed using a small number of commands
for any software. Technology assignments 6.6 to 6.8 provide the opportunity to learn more about the nor-
mal distribution, as well as the normal approximation to the binomial distribution.
TA6.1 Find the area under the standard normal curve
a. to the left of z = - 1.94 b. to the left of z = .83
c. to the right of z = 1.45 d. to the right of z = —1.65
e. between z = .75 and z = 1.90 f. between z = —1.20 and z = 1.55
TA6.2 Find the following areas under a normal curve with [x, = 86 and cr = 14.
a. Area to the left of x = 71 b. Area to the left of x = 96
c. Area to the right of x = 90 d. Area to the right of x = 75
e. Area between x = 65 and x = 75 f. Area between x = 72 and x = 95
TA6.3 The transmission on a particular model of car has a warranty for 40,000 miles. It is known that
the life of such a transmission has a normal distribution with a mean of 72,000 miles and a standard de-
viation of 12,000 miles. Answer the following questions.
a. What percentage of the transmissions will fail before the end of the warranty period?
b. What percentage of the transmissions will be good for more than 100,000 miles?
c. What percentage of the transmissions will be good for 80,000 to 100,000 miles?
TA6.4 Refer to Exercise 6.38. Assume that the distribution of time spent on leisure activities by cur-
rently employed adults living in households with no children younger than 18 years is normal with a mean
of 4.4 hours per day and a standard deviation of 1 .08 hours per day. Find the probability that the amount
of time spent on leisure activities per day for a randomly chosen person selected from the population of
interest (employed adults living in households with no children younger than 18 years) is
a. more than 7.2 hours per day
b. 4.2 to 6.5 hours per day
c. less than hours per day (theoretically, the normal distribution extends from negative infinity to pos-
itive infinity; realistically, time spent on leisure activity cannot be negative, so this answer provides
an idea of the level of approximation used in modeling this variable).
d. more than 24 hours per day (this is similar to part c, except that we are looking at the upper tail of
the distribution).
e. How much time must be spent on leisure activities by an employed adult living in households with
no children younger than 18 years to be in the group of such adults who spend the highest 3.5% of
time in a day on such activities?
TA6.5 Refer to Exercise 6.39. Suppose that the current percentage of after-tax income spent on food by all
U.S. residents aged 45 to 54 years follows a normal distribution with a mean of 9.32% and a standard devia-
tion of 1.38%. Find the proportion of such persons whose percentage of after-tax income spent on food is
a. less than 7.0%
b. between 10.0% and 12.5%.
c. A 47-year old U.S. resident with an after-tax income of $61,300 spent $5,485 on food last year. What
proportion of all U.S. residents aged 45 to 54 years spent a percentage of their after-tax income on
food that is less than that of this person?
TA6.6 In this assignment, we describe a method for determining how well a data set matches a normal
distribution. To do this, perform the following steps.
a. Using technology, arrange your data in increasing order.
b. In a separate column, enter the value of the quantile that each observation represents. To determine
the quantile, perform the following steps:
i. Calculate j r For example, if there are 5 data points, each observation divides the distribution into
groups of 20% (or .20). If there are 10 data points, each observation divides the distribution into
groups of 10% (or .10).
ii. The quantile corresponding to the smallest observation is equal to half of the result in part a. If
there are 5 data points, the quantile corresponding to the smallest observation is .10 (from .2/2).
Technology Assignments 299
If there are 10 data points, the quantile corresponding to the smallest observation is .05 (from
.1/2). The remaining quantiles can be calculated by repetitively adding the value - to the quantile
corresponding to the smallest observation. For example, if there are 5 data points, the quantiles
corresponding to the observations are .10, .30, .50, .70, and .90 (note that the gaps between the
quantiles are all equal to .20). If there are 10 data points, the quantiles corresponding to the ob-
servations are .05, .15, .25, .35, .45, .55, .65, .75, .85, and .95 (note that the gaps between the
quantiles are all equal to .10). Enter the corresponding quantile in the column next to its corre-
sponding observation.
c. Calculate the z values of the standard normal distribution corresponding to the quantiles determined
in step b-ii. Enter those values into another column, in the same rows as their corresponding quantiles.
d. Create a two-dimensional plot (called a scatterpJot) by placing the data on the horizontal axis and the
z scores that you derived in step c on the vertical axis. The command for creating this plot will
depend on the statistical software you are using. The commands for the three technologies presented
in this book can be found in the Technology Instruction section of Chapter 13.
If the distribution of your data matches the normal distribution exactly, the scatterplot will result in a di-
agonal line that moves from the lower left to the upper right. As your data differ more and more from a
normal distribution, the scatterplot will differ more and more from the diagonal line.
Follow the preceding steps to create the corresponding plot for the following data set.
185 188 190 195 210 218 225 225 234
Does the plot appear to be approximately linear? If not, how does it differ?
TA6.7 Technology assignment TA6.6 gave the instructions for creating a normal quantile plot, which can
be used to assess how well a data set matches a normal distribution. The use of a normal quantile plot is very
important in statistical inference procedures that require that the data come from a normal distribution. Statis-
tical software packages, as well as the TI graphing calculator, contain commands for making normal quantile
plots, whereas spreadsheet packages, such as Excel, require the user to follow the commands given in TA6.6.
Use your technology of choice to create a normal quantile plot and a histogram or dotplot for each of the fol-
lowing data sets. After creating the plots, make a conjecture to determine how the various shapes of a normal
quantile plot correspond to the shape characteristics (symmetry, skewness, outliers) of a distribution.
a. Exercise 3.71
b. Exercise 3.105
c. Exercise 3.136 without the two largest values
d. Exercise 3.136 without the three largest values
e. Exercise 3.140
TA6.8 As noted in Exercise 6.73, 4.8% of U.S. households in 2007 were classified as "linguistically iso-
lated," which means that all members of the household aged 14 and older have difficulty speaking Eng-
lish. Assume that this percentage is true for the current population of U.S. households. For each part of
that problem, calculate the requested probabilities using the following three methods:
i. Calculate the exact probability using the binomial distribution.
ii. Calculate the approximate probability using the normal distribution without the continuity correction.
iii. Calculate the approximate probability using the normal distribution with the continuity correction.
a. Using a sample size of 75, calculate the probability that 5 or more households are classified as lin-
guistically isolated.
b. Using a sample size of 150, calculate the probability that 10 or more households are classified as lin-
guistically isolated.
c. Using a sample size of 300, calculate the probability that 20 or more households are classified as lin-
guistically isolated.
d. Using a sample size of 600, calculate the probability that 40 or more households are classified as lin-
guistically isolated.
e. For each of parts a to d, what is the ratio of the number of households classified as linguistically iso-
lated to the sample size?
f. Comment on the relationship between the exact probability and the approximate probability using the
continuity correction as the sample size increases. Does the same relationship exist when the conti-
nuity correction is not used?
Chapter
Sampling Distributions
7.1 Population and
Sampling Distributions
7.2 Sampling and
Nonsampling Errors
7.3 Mean and Standard
Deviation of x
7.4 Shape of the Sampling
Distribution of x
7.5 Applications of the
Sampling Distribution
of x
7.6 Population and Sample
Proportions
7.7 Mean, Standard
Deviation, and Shape of
the Sampling
Distribution of p
7.8 Applications of the
Sampling Distribution
of p
You read about opinion polls in newspapers, magazines, and on the web every day. These polls
are based on sample surveys. Have you heard of sampling and nonsampling errors? It is good to
be aware of such errors while reading these opinion poll results. Sound sampling methods are
essential for opinion poll results to be valid and to lower the effects of such errors.
Chapters 5 and 6 discussed probability distributions of discrete and continuous random variables. This
chapter extends the concept of probability distribution to that of a sample statistic. As we discussed in
Chapter 3, a sample statistic is a numerical summary measure calculated for sample data. The mean,
median, mode, and standard deviation calculated for sample data are called sample statistics. On the
other hand, the same numerical summary measures calculated for population data are called popula-
tion parameters. A population parameter is always a constant, whereas a sample statistic is always a ran-
dom variable. Because every random variable must possess a probability distribution, each sample sta-
tistic possesses a probability distribution. The probability distribution of a sample statistic is more commonly
called its sampling distribution. This chapter discusses the sampling distributions of the sample mean
and the sample proportion. The concepts covered in this chapter are the foundation of the inferential
statistics discussed in succeeding chapters.
300
7.1 Population and Sampling Distributions
301
7.1 Population and Sampling
Distributions
This section introduces the concepts of population distribution and sampling distribution. Sub-
section 7.1.1 explains the population distribution, and Subsection 7.1.2 describes the sampling
distribution of 3c.
7.1.1 Population Distribution
The population distribution is the probability distribution derived from the information on all
elements of a population.
Definition
Population Distribution The population distribution is the probability distribution of the popu-
lation data.
Suppose there are only five students in an advanced statistics class and the midterm scores
of these five students are
70 78 80 80 95
Let x denote the score of a student. Using single-valued classes (because there are only five data
values, there is no need to group them), we can write the frequency distribution of scores as in
Table 7.1 along with the relative frequencies of classes, which are obtained by dividing the fre-
quencies of classes by the population size. Table 7.2, which lists the probabilities of various
x values, presents the probability distribution of the population. Note that these probabilities are
the same as the relative frequencies.
Table 7.1
Population Frequency and
Relative Frequency Distributions
Table 7.2
Population
Probability
Relative
Distribution
X
/
Frequency
X
P(x)
70
1
1/5 = .20
70
.20
78
1
1/5 = .20
78
.20
80
2
2/5 = .40
80
.40
95
1
1/5 = .20
95
.20
N = 5
Sum = 1.00
%P(x) = 1.00
The values of the mean and standard deviation calculated for the probability distribu-
tion of Table 7.2 give the values of the population parameters /jl and cr. These values are
pi = 80.60 and cr = 8.09. The values of p and cr for the probability distribution of Table 7.2
can be calculated using the formulas given in Sections 5.3 and 5.4 of Chapter 5 (see Exer-
cise 7.6).
7.1.2 Sampling Distribution
As mentioned at the beginning of this chapter, the value of a population parameter is al-
ways constant. For example, for any population data set, there is only one value of the
302 Chapter 7 Sampling Distributions
population mean, j±. However, we cannot say the same about the sample mean, x. We would
expect different samples of the same size drawn from the same population to yield differ-
ent values of the sample mean, x. The value of the sample mean for any one sample will
depend on the elements included in that sample. Consequently, the sample mean, x, is a ran-
dom variable. Therefore, like other random variables, the sample mean possesses a proba-
bility distribution, which is more commonly called the sampling distribution of x. Other
sample statistics, such as the median, mode, and standard deviation, also possess sampling
distributions.
Definition
Sampling Distribution of x The probability distribution of x is called its sampling
distribution. It lists the various values that x can assume and the probability of each value
of x.
In general, the probability distribution of a sample statistic is called its sampling
distribution.
Reconsider the population of midterm scores of five students given in Table 7.1. Consider
all possible samples of three scores each that can be selected, without replacement, from that
population. The total number of possible samples, given by the combinations formula discussed
in Chapter 5, is 10; that is,
5! 5 • 4 • 3 • 2 • 1
Total number of samples = = — ; — = =10
3!(5 - 3)! 3 • 2 • 1 • 2 • 1
Suppose we assign the letters A, B, C, D, and E to the scores of the five students, so that
A = 70, B = 78, C = 80, D = 80, E = 95
Then, the 10 possible samples of three scores each are
ABC, ABD, ABE, ACD, ACE, ADE, BCD, BCE, BDE, CDE
These 10 samples and their respective means are listed in Table 7.3. Note that the first
two samples have the same three scores. The reason for this is that two of the students
(C and D) have the same score, and, hence, the samples ABC and ABD contain the same
values. The mean of each sample is obtained by dividing the sum of the three scores in-
cluded in that sample by 3. For instance, the mean of the first sample is (70 + 78 + 80)/3 =
76. Note that the values of the means of samples in Table 7.3 are rounded to two decimal
places.
By using the values of x given in Table 7.3, we record the frequency distribution of x in
Table 7.4. By dividing the frequencies of the various values of x by the sum of all frequencies,
we obtain the relative frequencies of classes, which are listed in the third column of Table 7.4.
These relative frequencies are used as probabilities and listed in Table 7.5. This table gives the
sampling distribution of x.
If we select just one sample of three scores from the population of five scores, we may
draw any of the 10 possible samples. Hence, the sample mean, x, can assume any of the val-
ues listed in Table 7.5 with the corresponding probability. For instance, the probability that
the mean of a randomly selected sample of three scores is 81.67 is .20. This probability can
be written as
P{x = 81.67) = .20
7.2 Sampling and Nonsampling Errors 303
la hi a "IX
IdDIc I.J
All Possible Samples and
Their Means When the
Takln 7/1
IdDIc /.4
Frequency and Relative
Frequency Distributions of x
la hi a 7 R
let Die /.D
Sampling Distri-
bution of X
Sample Size Is 3
When the Sample Size Is 3
When the Sam-
Scores in
Relative
ple Size Is 3
Sample
the Sample
X
X
/
Frequency
.V
P(x)
AdL
70, 78, 80
76.00
/O.UU
2
2/10 =
.20
/O.UU
./A)
a on
70, 78, 80
76.00
/O.O /
1
1/10 =
.10
/O.O /
A DP
A D Ll
70, 78, 95
81.00
7Q 11.
/y.JJ
1
1/10 =
.10
/y.jj
. 1U
ACD
70, 80, 80
76.67
81.00
1
1/10 =
.10
81.00
.10
70, 80, 95
81.67
ol .0 /
2
2/10 =
.20
Q 1 A 7
.ZAJ
ADE
70, 80, 95
81.67
84. 11
2
2/10 =
.20
Szl 11
.—\)
BCD
78, 80, 80
79.33
85.00
1
1/10 =
.10
85.00
.10
BCE
78, 80, 95
84.33
2/= 10
Sum =
1.00
%P(x) = 1.00
BDE
78, 80, 95
84.33
CDE
80, 80, 95
85.00
7.2 Sampling and Nonsampling Errors
Usually, different samples selected from the same population will give different results because
they contain different elements. This is obvious from Table 7.3, which shows that the mean of
a sample of three scores depends on which three of the five scores are included in the sample.
The result obtained from any one sample will generally be different from the result obtained
from the corresponding population. The difference between the value of a sample statistic ob-
tained from a sample and the value of the corresponding population parameter obtained from
the population is called the sampling error. Note that this difference represents the sampling
error only if the sample is random and no nonsampling error has been made. Otherwise, only
a part of this difference will be due to the sampling error.
Definition
Sampling Error Sampling error is the difference between the value of a sample statistic and the
value of the corresponding population parameter. In the case of the mean,
Sampling error = x — fi
assuming that the sample is random and no nonsampling error has been made.
It is important to remember that a sampling error occurs because of chance. The errors that
occur for other reasons, such as errors made during collection, recording, and tabulation of data,
are called nonsampling errors. These errors occur because of human mistakes, and not chance.
Note that there is only one kind of sampling error — the error that occurs due to chance. How-
ever, there is not just one nonsampling error, but there are many nonsampling errors that may
occur for different reasons.
Definition
Nonsampling Errors The errors that occur in the collection, recording, and tabulation of data are
called nonsampling errors.
304 Chapter 7 Sampling Distributions
The following paragraph, reproduced from the Current Population Reports of the U.S. Bu-
reau of the Census, explains how nonsampling errors can occur.
Nonsampling errors can be attributed to many sources, e.g., inability to obtain information about
all cases in the sample, definitional difficulties, differences in the interpretation of questions, in-
ability or unwillingness on the part of the respondents to provide correct information, inability to
recall information, errors made in collection such as in recording or coding the data, errors made
in processing the data, errors made in estimating values for missing data, biases resulting from
the differing recall periods caused by the interviewing pattern used, and failure of all units in the
universe to have some probability of being selected for the sample (undercoverage).
The following are the main reasons for the occurrence of nonsampling errors.
1. If a sample is nonrandom (and, hence, nonrepresentative), the sample results may be too
different from the census results. The following quote from U.S. News & World Report
describes how even a randomly selected sample can become nonrandom if some of the
members included in the sample cannot be contacted.
A test poll conducted in the 1984 presidential election found that if the poll were halted after inter-
viewing only those subjects who could be reached on the first try, Reagan showed a 3-percentage-
point lead over Mondale. But when interviewers made a determined effort to reach everyone on
their lists of randomly selected subjects — calling some as many as 30 times before finally reach-
ing them — Reagan showed a 13 percent lead, much closer to the actual election result. As it
turned out, people who were planning to vote Republican were simply less likely to be at home.
("The Numbers Racket: How Polls and Statistics Lie," U.S. News & World Report, July 11, 1988.
Copyright © 1988 by U.S. News & World Report, Inc. Reprinted with permission.)
2. The questions may be phrased in such a way that they are not fully understood by
the members of the sample or population. As a result, the answers obtained are not accurate.
3. The respondents may intentionally give false information in response to some sensitive ques-
tions. For example, people may not tell the truth about their drinking habits, incomes, or
opinions about minorities. Sometimes the respondents may give wrong answers because of
ignorance. For example, a person may not remember the exact amount he or she spent on
clothes during the last year. If asked in a survey, he or she may give an inaccurate answer.
4. The poll taker may make a mistake and enter a wrong number in the records or make an
error while entering the data on a computer.
Note that nonsampling errors can occur both in a sample survey and in a census, whereas
sampling error occurs only when a sample survey is conducted. Nonsampling errors can be
minimized by preparing the survey questionnaire carefully and handling the data cautiously.
However, it is impossible to avoid sampling error.
Example 7-1 illustrates the sampling and nonsampling errors using the mean.
Illustrating sampling
and nonsampling errors.
■ EXAMPLE 7-1
Reconsider the population of five scores given in Table 7.1. Suppose one sample of three scores
is selected from this population, and this sample includes the scores 70, 80, and 95. Find the
sampling error.
Solution The scores of the five students are 70, 78, 80, 80, and 95. The population mean is
70 + 78 + 80 + 80 + 95
/x = = 80.60
Now a random sample of three scores from this population is taken and this sample includes
the scores 70, 80, and 95. The mean for this sample is
_ 70 + 80 + 95
x = = 81.67
3
7.2 Sampling and Nonsampling Errors 305
Consequently,
Sampling error = x - (jl = 81.67 - 80.60 = 1.07
That is, the mean score estimated from the sample is 1 .07 higher than the mean score of the
population. Note that this difference occurred due to chance — that is, because we used a sam-
ple instead of the population. H
Now suppose, when we select the sample of three scores, we mistakenly record the second
score as 82 instead of 80. As a result, we calculate the sample mean as
_ 70 + 82 + 95
x = = 82.33
3
Consequently, the difference between this sample mean and the population mean is
x - fi = 82.33 - 80.60 = 1.73
However, this difference between the sample mean and the population mean does not represent
the sampling error. As we calculated earlier, only 1 .07 of this difference is due to the sampling
error. The remaining portion, which is equal to 1.73 — 1 .07 = .66, represents the nonsampling
error because it occurred due to the error we made in recording the second score in the sam-
ple. Thus, in this case,
Samplinge rror = 1.07
Nonsamplinge rror = .66
Figure 7.1 shows the sampling and nonsampling errors for these calculations.
Sampling error Nonsampling error
A .. A
I 1 1
H = 80.60 81.67 82.33
Figure 7.1 Sampling and nonsampling errors.
Thus, the sampling error is the difference between the correct value of x and /a, where the
correct value of x is the value of x that does not contain any nonsampling errors. In contrast,
the nonsampling error(s) is (are) obtained by subtracting the correct value of x from the incor-
rect value of x, where the incorrect value of x is the value that contains the nonsampling error(s).
For our example,
Samplinge rror = x — fi = 81.67 - 80.60 = 1.07
Nonsamplinge rror = Incorrect x — Correct x = 82.33 — 81.67 = .66
Note that in the real world we do not know the mean of a population. Hence, we select a
sample to use the sample mean as an estimate of the population mean. Consequently, we never
know the size of the sampling error.
EXERCISES
CONCEPTS AND PROCEDURES
7.1 Briefly explain the meaning of a population distribution and a sampling distribution. Give an exam-
ple of each.
7.2 Explain briefly the meaning of sampling error. Give an example. Does such an error occur only in a
sample survey, or can it occur in both a sample survey and a census?
7.3 Explain briefly the meaning of nonsampling errors. Give an example. Do such errors occur only in
a sample survey, or can they occur in both a sample survey and a census?
306
Chapter 7 Sampling Distributions
7.4 Consider the following population of six numbers.
1 51 38 1 79 1 2
a. Find the population mean.
b. Liza selected one sample of four numbers from this population. The sample included the numbers
13, 8, 9, and 12. Calculate the sample mean and sampling error for this sample.
c. Refer to part b. When Liza calculated the sample mean, she mistakenly used the numbers 13, 8, 6,
and 12 to calculate the sample mean. Find the sampling and nonsampling errors in this case.
d. List all samples of four numbers (without replacement) that can be selected from this population.
Calculate the sample mean and sampling error for each of these samples.
7.5 Consider the following population of 10 numbers.
2 02 51 31 99 1 51 17 1 73
a. Find the population mean.
b. Rich selected one sample of nine numbers from this population. The sample included the num-
bers 20, 25, 13, 9, 15, 11, 7, 17, and 30. Calculate the sample mean and sampling error for this
sample.
c. Refer to part b. When Rich calculated the sample mean, he mistakenly used the numbers 20, 25,
13, 9, 15, 11, 17, 17, and 30 to calculate the sample mean. Find the sampling and nonsampling
errors in this case.
d. List all samples of nine numbers (without replacement) that can be selected from this population.
Calculate the sample mean and sampling error for each of these samples.
■ APPLICATIONS
7.6 Using the formulas of Sections 5.3 and 5.4 of Chapter 5 for the mean and standard deviation of a
discrete random variable, verify that the mean and standard deviation for the population probability dis-
tribution of Table 7.2 are 80.60 and 8.09, respectively.
7.7 The following data give the ages (in years) of all six members of a family.
55 53 28 25 21 15
a. Let x denote the age of a member of this family. Write the population distribution of x.
b. List all the possible samples of size five (without replacement) that can be selected from this
population. Calculate the mean for each of these samples. Write the sampling distribution of x.
c. Calculate the mean for the population data. Select one random sample of size five and calculate
the sample mean x. Compute the sampling error.
7.8 The following data give the years of teaching experience for all five faculty members of a department
at a university.
7 8 1 47 2
a. Let x denote the years of teaching experience for a faculty member of this department. Write the
population distribution of x.
b. List all the possible samples of size four (without replacement) that can be selected from this
population. Calculate the mean for each of these samples. Write the sampling distribution of x.
c. Calculate the mean for the population data. Select one random sample of size four and calculate
the sample mean x. Compute the sampling error.
7.3 Mean and Standard Deviation of x
The mean and standard deviation calculated for the sampling distribution of x are called the
mean and standard deviation of x. Actually, the mean and standard deviation of x are, re-
spectively, the mean and standard deviation of the means of all samples of the same size se-
lected from a population. The standard deviation of x is also called the standard error of x.
Definition
Mean and Standard Deviation of x The mean and standard deviation of the sampling distribution
of x are called the mean and standard deviation ofx and are denoted by y^. and cr^, respectively.
7.3 Mean and Standard Deviation of x 307
If we calculate the mean and standard deviation of the 10 values of x listed in Table 7.3,
we obtain the mean, /jl^, and the standard deviation, o^, of x. Alternatively, we can calculate the
mean and standard deviation of the sampling distribution of x listed in Table 7.5. These will
also be the values of \xj, and <j- x . From these calculations, we will obtain p.- x = 80.60 and
cr- x = 3.30 (see Exercise 7.25 at the end of this section).
The mean of the sampling distribution of x is always equal to the mean of the population.
Mean of the Sampling Distribution of x The mean of the sampling distribution of x is always
equal to the mean of the population. Thus,
Hence, if we select all possible samples (of the same size) from a population and cal-
culate their means, the mean (p. x ) of all these sample means will be the same as the mean
(fi) of the population. If we calculate the mean for the population probability distribution of
Table 7.2 and the mean for the sampling distribution of Table 7.5 by using the formula learned
in Section 5.3 of Chapter 5, we get the same value of 80.60 for /jl and /jl x (see Exercise 7.25).
The sample mean, i, is called an estimator of the population mean, /jl. When the expected
value (or mean) of a sample statistic is equal to the value of the corresponding population parame-
ter, that sample statistic is said to be an unbiased estimator. For the sample mean x, ^ = p.. Hence,
x is an unbiased estimator of /jl. This is a very important property that an estimator should possess.
However, the standard deviation, cr s , of x is not equal to the standard deviation, cr, of the
population distribution (unless n = 1). The standard deviation of x is equal to the standard de-
viation of the population divided by the square root of the sample size; that is,
Vn
This formula for the standard deviation of x holds true only when the sampling is done either
with replacement from a finite population or with or without replacement from an infinite pop-
ulation. These two conditions can be replaced by the condition that the above formula holds
true if the sample size is small in comparison to the population size. The sample size is consid-
ered to be small compared to the population size if the sample size is equal to or less than 5%
of the population size — that is, if
n
- < .05
N
If this condition is not satisfied, we use the following formula to calculate cr^:
o- N
Vn \ N — I
IN - n
where the factor A — is called the finite population correction factor.
In most practical applications, the sample size is small compared to the population size.
Consequently, in most cases, the formula used to calculate a- x is Oj = cr/Vn.
Standard Deviation of the Sampling Distribution of x The standard deviation of the sampling
distribution of x is
cr
o-j = — 7=
vn
where cr is the standard deviation of the population and n is the sample size. This formula is
used when n/N ^ .05, where is the population size.
308 Chapter 7 Sampling Distributions
Following are two important observations regarding the sampling distribution of x.
1. The spread of the sampling distribution of x is smaller than the spread of the correspon-
ding population distribution. In other words, cr- < cr. This is obvious from the formula for
a x . When n is greater than 1, which is usually true, the denominator in cr/Vn is greater
than 1 . Hence, a x is smaller than cr.
2. The standard deviation of the sampling distribution of x decreases as the sample size in-
creases. This feature of the sampling distribution of x is also obvious from the formula
cr
°"* = — 7=
Vn
If the standard deviation of a sample statistic decreases as the sample size is increased, that
statistic is said to be a consistent estimator. This is another important property that an estimator
should possess. It is obvious from the above formula for a"j that as n increases, the value of Vn
also increases and, consequently, the value of cr/Vn decreases. Thus, the sample mean x is a con-
sistent estimator of the population mean jx. Example 7-2 illustrates this feature.
Finding the mean and
standard deviation ofx.
■ EXAMPLE 7-2
The mean wage per hour for all 5000 employees who work at a large company is $27.50, and
the standard deviation is $3.70. Let x be the mean wage per hour for a random sample of cer-
tain employees selected from this company. Find the mean and standard deviation of x for a
sample size of
(a) 30
(b) 75
(c) 200
(a)
Solution From the given information, for the population of all employees,
N = 5000, {jl = $27.50, and a = $3.70
The mean, /j.^, of the sampling distribution of x is
/jl x = fji = $27.50
In this case, n = 30, N = 5000, and n/N = 30/5000 = .006. Because n/N is less
than .05, the standard deviation of x is obtained by using the formula cr/Vn. Hence,
cr _ 170
a " ~ Vn ~ V30
(b)
(c)
$.676
Thus, we can state that if we take all possible samples of size 30 from the population
of all employees of this company and prepare the sampling distribution of x, the mean
and standard deviation of this sampling distribution of x will be $27.50 and $.676,
respectively.
In this case, n = 75 and n/N = 75/5000 =
and standard deviation of x are
/U/j = fi = $27.50 and cr T = -
In this case, n = 200 and n/N = 200/5000
the mean and standard deviation of x are
.015, which is less than .05. The mean
3.70
$.427
.04, which is less than .05. Therefore,
fi x = /jl = $27.50 and cr-
cr
Vn
3.70
V200
$.262
From the preceding calculations we observe that the mean of the sampling distribu-
tion of x is always equal to the mean of the population whatever the size of the sam-
ple. However, the value of the standard deviation of x decreases from $.676 to $.427 and
then to $.262 as the sample size increases from 30 to 75 and then to 200. B
7.3 Mean and Standard Deviation of x
309
EXERCISES
CONCEPTS AND PROCEDURES
7.9 Let x be the mean of a sample selected from a population.
a. What is the mean of the sampling distribution of x equal to?
b. What is the standard deviation of the sampling distribution of x equal to? Assume n/N ^ .05.
7.10 What is an estimator? When is an estimator unbiased? Is the sample mean, x, an unbiased estimator
of /jl7 Explain.
7.11 When is an estimator said to be consistent? Is the sample mean, x, a consistent estimator of \xP. Explain.
7.12 How does the value of <j- x change as the sample size increases? Explain.
7.13 Consider a large population with = 60 and cr = 10. Assuming n/N — .05, find the mean and stan-
dard deviation of the sample mean, x, for a sample size of
a. 18 b. 90
7.14 Consider a large population with /jl = 90 and cr = 18. Assuming n/N < .05, find the mean and stan-
dard deviation of the sample mean, x, for a sample size of
a. 10 b. 35
7.15 A population of N = 5000 has cr = 25. In each of the following cases, which formula will you use
to calculate cr T and why? Using the appropriate formula, calculate cr T for each of these cases.
a. n = 300 b. n = 100
7.16 A population of N = 100,000 has cr = 40. In each of the following cases, which formula will you
use to calculate cr^ and why? Using the appropriate formula, calculate cr^ for each of these cases.
a. n = 2500 b. n = 7000
*7.17 For a population, fi = 125 and cr = 36.
a. For a sample selected from this population, /x Y = 125 and cr- t = 3.6. Find the sample size.
Assume n/N < .05.
b. For a sample selected from this population, ^ = 125 and crj. = 2.25. Find the sample size.
Assume n/N < .05.
*7.18 For a population, /x = 46 and cr = 10.
a. For a sample selected from this population, ^ = 46 and cr^ = 2.0. Find the sample size.
Assume n/N < .05.
b. For a sample selected from this population, /u,j = 46 and cr^ = 1.6. Find the sample size.
Assume n/N < .05.
7.19 According to the University of Wisconsin Dairy Marketing and Risk Management Program, the av-
erage retail price of a gallon of whole milk in the United States for April 2009 was $3,084
(http://future.aae.wisc.edu/index.html). Suppose that the current distribution of the retail prices of a gal-
lon of whole milk in the United States has a mean of $3,084 and a standard deviation of $.263. Let x be
the average retail price of a gallon of whole milk for a random sample of 47 stores. Find the mean and
the standard deviation of the sampling distribution of x.
7.20 The living spaces of all homes in a city have a mean of 2300 square feet and a standard deviation
of 500 square feet. Let x be the mean living space for a random sample of 25 homes selected from this
city. Find the mean and standard deviation of the sampling distribution of x.
7.21 The mean monthly out-of-pocket cost of prescription drugs for all senior citizens in a particular city
is $520 with a standard deviation of $72. Let x be the mean of such costs for a random sample of 25 sen-
ior citizens from this city. Find the mean and standard deviation of the sampling distribution of x.
7.22 An article in the Daily Herald of Everett, Washington, noted that the average cost of going to a mi-
nor league baseball game for a family of four was $55 in 2009 (http://www.heraldnet.com/article/
20090412/BIZ/704129929/1006/SPORTS03). Suppose that the standard deviation of such costs is $13.25.
Let x be the average cost of going to a minor league baseball game for 33 randomly selected families of
four in 2009. Find the mean and the standard deviation of the sampling distribution of x.
*7.23 Suppose the standard deviation of recruiting costs per player for all female basketball players
recruited by all public universities in the Midwest is $2000. Let x be the mean recruiting cost for a
■ APPLICATIONS
310
Chapter 7 Sampling Distributions
sample of a certain number of such players. What sample size will give the standard deviation of x
equal to $125?
*7.24 The standard deviation of the 2009 gross sales of all corporations is known to be $139.50 million.
Let x be the mean of the 2009 gross sales of a sample of corporations. What sample size will produce the
standard deviation of x equal to $15.50 million?
*7.25 Consider the sampling distribution of x given in Table 7.5.
a. Calculate the value of ^ using the formula ^ = 'XxPfe). Is the value of /x calculated in Exer-
cise 7.6 the same as the value of fa calculated here?
b. Calculate the value of <r- x by using the formula
o-j = Vxx 2 p(x) - (^y-
c. From Exercise 7.6, a = 8.09. Also, our sample size is 3, so that n = 3. Therefore,
a/Vn = 8.09/V3 = 4.67. From part b, you should get cr T = 3.30. Why does cr/Vn not equal
<xj in this case?
d. In our example (given in the beginning of Section 7.1.1) on scores, N = 5 and n = 3. Hence,
n/N = 3/5 = .60. Because n/N is greater than .05, the appropriate formula to find u- x is
a IN - n
a *~ Vf! V N - 1
Show that the value of cr T calculated by using this formula gives the same value as the one cal-
culated in part b above.
7.4 Shape of the Sampling Distribution of x
The shape of the sampling distribution of x relates to the following two cases.
1. The population from which samples are drawn has a normal distribution.
2. The population from which samples are drawn does not have a normal distribution.
7.4.1 Sampling from a Normally Distributed Population
When the population from which samples are drawn is normally distributed with its mean equal
to fi and standard deviation equal to o\ then:
1. The mean of x, /jl^, is equal to the mean of the population, /jl.
2. The standard deviation of x, Oj, is equal to a/Vn, assuming n/N ^ .05.
3. The shape of the sampling distribution of x is normal, whatever the value of n.
Sampling Distribution of x When the Population Has a Normal Distribution If the population
from which the samples are drawn is normally distributed with mean /jl and standard devia-
tion cr, then the sampling distribution of the sample mean, x, will also be normally distrib-
uted with the following mean and standard deviation, irrespective of the sample size:
O"
l±- = i± and a, = — 1=
Remember ► For cr^ = cr/Vn to be true, n/N must be less than or equal to .05.
Figure 7.2a shows the probability distribution curve for a population. The distribution
curves in Figure 1 .2b through Figure 7.2e show the sampling distributions of x for different
sample sizes taken from the population of Figure 7.2a. As we can observe, the population has
a normal distribution. Because of this, the sampling distribution of x is normal for each of
7.4 Shape of the Sampling Distribution of x 311
(d) Sampling distribution of x for n = 30.
(e) Sampling distribution of ,v for « = 1 00.
A Normal
\ distribution
X
\ Normal
J
\ distribution
the four cases illustrated in Figure 1 .2b through Figure 7.2e. Also notice from Figure 1 .2b
through Figure 7.2e that the spread of the sampling distribution of x decreases as the sample
size increases.
Example 7-3 illustrates the calculation of the mean and standard deviation of x and the de-
scription of the shape of its sampling distribution.
■ EXAMPLE 7-3
In a recent SAT, the mean score for all examinees was 1020. Assume that the distribution of SAT
scores of all examinees is normal with a mean of 1020 and a standard deviation of 153. Let x
be the mean SAT score of a random sample of certain examinees. Calculate the mean and stan-
dard deviation of x and describe the shape of its sampling distribution when the sample size is
(a) 16
(b) 50
(c) 1000
Finding the mean, standard
deviation, and sampling
distribution of x: normal
population.
Solution Let \x and a be the mean and standard deviation of SAT scores of all examinees,
and let yu-j and cr^ be the mean and standard deviation of the sampling distribution of x, respec-
tively. Then, from the given information,
fi, = 1020 and cr = 153
(a) The mean and standard deviation of x are, respectively,
a _ 153
1020 and
38.250
Because the SAT scores of all examinees are assumed to be normally distributed, the
sampling distribution of x for samples of 16 examinees is also normal. Figure 7.3
312 Chapter 7 Sampling Distributions
shows the population distribution and the sampling distribution of x. Note that because
cr is greater than cr-, the population distribution has a wider spread but smaller height
than the sampling distribution of x in Figure 7.3.
Figure 7.3
o> = 38.250 -W
W- Sampling distribution
of x for« =16
V"v r- Population
° = 1 53 ^I^^S
\^\ s i distribution
^ = fi = 1 020
SAT scores
(b) The mean and standard deviation of x are, respectively,
jjbx = jju = 1020 and <j- x
50
21.637
Again, because the SAT scores of all examinees are assumed to be normally distrib-
uted, the sampling distribution of x for samples of 50 examinees is also normal. The
population distribution and the sampling distribution of x are shown in Figure 7.4.
Figure 7.4
<7j = 21 .637 >■/
Y* Sampling distribution
\ of xforrc = 50
cr = 1 53 — | yy
VS. i- Population
\ s^l distribution
jUj = \x = 1 020 SAT scores
(c) The mean and standard deviation of x are, respectively,
a 153
ix.- = u = 1020 and <r T = — p = , = 4.838
Vn VTOOO
Again, because the SAT scores of all examinees are assumed to be normally distrib-
uted, the sampling distribution of x for samples of 1000 examinees is also normal.
The two distributions are shown in Figure 7.5.
Figure 7.5
cr- = 4.838 *\
\* Sampling distribution
1 of x for n =1000
o-= 153—1 >S J
\ ^ — p Population
\ distribution
fi- = n = 1 020
SAT scores
Thus, whatever the sample size, the sampling distribution of x is normal when
the population from which the samples are drawn is normally distributed. H
7.4 Shape of the Sampling Distribution of x
313
7.4.2 Sampling from a Population That Is Not
Normally Distributed
Most of the time the population from which the samples are selected is not normally distrib-
uted. In such cases, the shape of the sampling distribution of x is inferred from a very impor-
tant theorem called the central limit theorem.
Central Limit Theorem According to the central limit theorem, for a large sample size, the sam-
pling distribution of x is approximately normal, irrespective of the shape of the population dis-
tribution. The mean and standard deviation of the sampling distribution of x are, respectively,
a
fJ-i = M and a 7x = —=
Vn
The sample size is usually considered to be large if n ^ 30.
Note that when the population does not have a normal distribution, the shape of the sam-
pling distribution is not exactly normal, but it is approximately normal for a large sample size.
The approximation becomes more accurate as the sample size increases. Another point to
remember is that the central limit theorem applies to large samples only. Usually, if the sample
size is 30 or more, it is considered sufficiently large so that the central limit theorem can be
applied to the sampling distribution of x. Thus, according to the central limit theorem:
1. When n s 30, the shape of the sampling distribution of x is approximately normal irre-
spective of the shape of the population distribution.
2. The mean of x, \x- x , is equal to the mean of the population, fi.
3. The standard deviation of x, cr s , is equal to cr/Vn.
Again, remember that for = cr/Vn to apply, n/N must be less than or equal to .05.
Figure 7.6a shows the probability distribution curve for a population. The distribution curves
in Figure 7.6b through Figure 7.6e show the sampling distributions of x for different sample
(a) Population distribution.
{b) Sampling distribution of .y for n = 4.
(c) Sampling distribution of x for n = 1 5.
(d) Sampling distribution of x for n = 30.
(e) Sampling distribution of x for n = 80.
Figure 7.6 Population distribution and
X
X
X
Approximately
■\ normal
\ distribution
X
J
^ Approximately
\ normal
\ distribution
sampling distributions of x.
314
Chapter 7 Sampling Distributions
sizes taken from the population of Figure 7.6a. As we can observe, the population is not nor-
mally distributed. The sampling distributions of x shown in parts b and c, when n < 30, are not
normal. However, the sampling distributions of x shown in parts d and e, when n > 30, are (ap-
proximately) normal. Also notice that the spread of the sampling distribution of x decreases as
the sample size increases.
Example 7-4 illustrates the calculation of the mean and standard deviation of x and de-
scribes the shape of the sampling distribution of x when the sample size is large.
Finding the mean,
standard deviation, and
sampling distribution of
x: nonnormal population.
■ EXAMPLE 7-4
The mean rent paid by all tenants in a small city is $1550 with a standard deviation of $225.
However, the population distribution of rents for all tenants in this city is skewed to the right.
Calculate the mean and standard deviation of x and describe the shape of its sampling distri-
bution when the sample size is
(a) 30
(b) 100
Solution Although the population distribution of rents paid by all tenants is not normal, in
each case the sample size is large (n > 30). Hence, the central limit theorem can be applied
to infer the shape of the sampling distribution of x.
(a) Let x be the mean rent paid by a sample of 30 tenants. Then, the sampling distribution
of x is approximately normal with the values of the mean and standard deviation given as
fa = /jl = $1550 and cr- x
(T
Vn
225
30
$41,079
Figure 7.7 shows the population distribution and the sampling distribution of x.
= $1550 x /ij = $1550 x
(a) Population distribution. (b) Sampling distribution of x for n = 30.
Figure 7.7
(b) Let x be the mean rent paid by a sample of 100 tenants. Then, the sampling distribution
of x is approximately normal with the values of the mean and standard deviation given as
= $22,500
100
Figure 7.8 shows the population distribution and the sampling distribution of x.
/i = $1550 x juj = $1550 x
{a) Population distribution. (b) Sampling distribution of x for n = 100.
Figure 7.8
7.4 Shape of the Sampling Distribution of x 315
EXERCISES
CONCEPTS AND PROCEDURES
7.26 What condition or conditions must hold true for the sampling distribution of the sample mean to be
normal when the sample size is less than 30?
7.27 Explain the central limit theorem.
7.28 A population has a distribution that is skewed to the left. Indicate in which of the following cases
the central limit theorem will apply to describe the sampling distribution of the sample mean.
a. n = 400 b. n = 25 c. n = 36
7.29 A population has a distribution that is skewed to the right. A sample of size n is selected from this pop-
ulation. Describe the shape of the sampling distribution of the sample mean for each of the following cases.
a. n = 25 b. n = 80 c. n = 29
7.30 A population has a normal distribution. A sample of size n is selected from this population. Describe
the shape of the sampling distribution of the sample mean for each of the following cases.
a. n = 94 b. n = 11
7.31 A population has a normal distribution. A sample of size n is selected from this population. De-
scribe the shape of the sampling distribution of the sample mean for each of the following cases.
a. n = 23 b. n = 450
■ APPLICATIONS
7.32 The delivery times for all food orders at a fast-food restaurant during the lunch hour are normally
distributed with a mean of 7.7 minutes and a standard deviation of 2.1 minutes. Let x be the mean deliv-
ery time for a random sample of 16 orders at this restaurant. Calculate the mean and standard deviation
of x, and describe the shape of its sampling distribution.
7.33 Among college students who hold part-time jobs during the school year, the distribution of the time
spent working per week is approximately normally distributed with a mean of 20.20 hours and a standard
deviation of 2.60 hours. Let x be the average time spent working per week for 18 randomly selected col-
lege students who hold part-time jobs during the school year. Calculate the mean and the standard devia-
tion of the sampling distribution of x, and describe the shape of this sampling distribution.
7.34 The amounts of electricity bills for all households in a particular city have an approximately normal
distribution with a mean of $140 and a standard deviation of $30. Let x be the mean amount of electric-
ity bills for a random sample of 25 households selected from this city. Find the mean and standard devi-
ation of x, and comment on the shape of its sampling distribution.
7.35 The GPAs of all 5540 students enrolled at a university have an approximately normal distribution
with a mean of 3.02 and a standard deviation of .29. Let x be the mean GPA of a random sample of
48 students selected from this university. Find the mean and standard deviation of x, and comment on the
shape of its sampling distribution.
7.36 The weights of all people living in a particular town have a distribution that is skewed to the right
with a mean of 133 pounds and a standard deviation of 24 pounds. Let x be the mean weight of a random
sample of 45 persons selected from this town. Find the mean and standard deviation of x and comment
on the shape of its sampling distribution.
7.37 In an article by Laroche et al. (The Journal of the American Board of Family Medicine 2007;20:9-15),
the average daily fat intake of U.S. adults with children in the household is 91.4 grams, with a standard
deviation of 93.25 grams. These results are based on a sample of 3714 adults. Suppose that these results
hold true for the current population distribution of daily fat intake of such adults, and that this distribu-
tion is strongly skewed to the right. Let x be the average daily fat intake of 20 randomly selected U.S.
adults with children in the household. Find the mean and the standard deviation of the sampling distribu-
tion of x. Do the same for a random sample of size 75. How do the shapes of the sampling distributions
differ for the two sample sizes?
7.38 Suppose the incomes of all people in the United States who own hybrid (gas and electric) automo-
biles are normally distributed with a mean of $78,000 and a standard deviation of $8300. Let x be the
mean income of a random sample of 50 such owners. Calculate the mean and standard deviation of x and
describe the shape of its sampling distribution.
316
Chapter 7 Sampling Distributions
7.39 Annual per capita (average per person) chewing gum consumption in the United States is 200 pieces
(http://www.iplcricketlive.com/). Suppose that the standard deviation of per capita consumption is 145
pieces per year. Let x be the average annual chewing gum consumption of 84 randomly selected Americans.
Find the mean and the standard deviation of the sampling distribution of x. What is the shape of the sam-
pling distribution of x? Do you need to know the shape of the population distribution to make this con-
clusion? Explain why or why not.
7.5 Applications of the Sampling
Distribution of x
From the central limit theorem, for large samples, the sampling distribution of x is approxi-
mately normal with mean p and standard deviation <r- x = cr/Vn. Based on this result, we can
make the following statements about x for large samples. The areas under the curve of x men-
tioned in these statements are found from the normal distribution table.
1. If we take all possible samples of the same (large) size from a population and calculate
the mean for each of these samples, then about 68.26% of the sample means will be within
one standard deviation of the population mean. Alternatively, we can state that if we take
one sample (of n > 30) from a population and calculate the mean for this sample, the
probability that this sample mean will be within one standard deviation of the population
mean is .6826. That is,
P((jl - la 7x < x < p. + lo- T ) = .8413 - .1587 = .6826
This probability is shown in Figure 7.9.
Figure 7.9 - loj < x < /x, + laj)
2. If we take all possible samples of the same (large) size from a population and calculate
the mean for each of these samples, then about 95.44% of the sample means will be within
two standard deviations of the population mean. Alternatively, we can state that if we take
one sample (of n > 30) from a population and calculate the mean for this sample, the
probability that this sample mean will be within two standard deviations of the population
mean is .9544. That is,
P(p - 2o- 7x < x < p + 2o-j) = .9772 - .0228 = .9544
This probability is shown in Figure 7.10.
3. If we take all possible samples of the same (large) size from a population and calculate the
mean for each of these samples, then about 99. 74% of the sample means will be within
7.5 Applications of the Sampling Distribution of x 317
three standard deviations of the population mean. Alternatively, we can state that if we take
one sample (of n > 30) from a population and calculate the mean for this sample, the
probability that this sample mean will be within three standard deviations of the popula-
tion mean is .9974. That is,
P{l± - 3cr s < x < fi + 3o-£) = .9987 - .0013 = .9974
This probability is shown in Figure 7.11.
Ii- 3a- /J / j + 3 <t j x
When conducting a survey, we usually select one sample and compute the value of x based
on that sample. We never select all possible samples of the same size and then prepare the sam-
pling distribution of x. Rather, we are more interested in finding the probability that the value
of x computed from one sample falls within a given interval. Examples 7-5 and 7-6 illustrate
this procedure.
EXAMPLE 7-5
Assume that the weights of all packages of a certain brand of cookies are normally distrib-
. „ . . „ „. .,. Calculating the probability
uted with a mean of 32 ounces and a standard deviation of .3 ounce. Find the probability that ,_ . , ,
OJ X in an interval: normal
the mean weight, x, of a random sample of 20 packages of this brand of cookies will be be- population
tween 31.8 and 31.9 ounces.
Solution Although the sample size is small (n < 30), the shape of the sampling distribu-
tion of x is normal because the population is normally distributed. The mean and standard
deviation of x are, respectively,
(1%
jji = 32 ounces and cr Y
Vn V20
.06708204 ounce
We are to compute the probability that the value of x calculated for one randomly drawn sam-
ple of 20 packages is between 31.8 and 31.9 ounces — that is,
P(31.8 < x < 31.9)
This probability is given by the area under the normal distribution curve for x between the
points x = 31.8 and x = 31.9. The first step in finding this area is to convert the two x val-
ues to their respective z values.
1 Value for a Value of X The z value for a value of x is calculated as
x — j±
O-T.
The z values for x = 31.8 and x = 31.9 are computed next, and they are shown on the z scale
below the normal distribution curve for x in Figure 7.12.
31.8 - 32
For* = 31.8: z = =-2.98
.06708204
31.9 - 32
For* = 31.9: z = =-1.49
.06708204
318 Chapter 7 Sampling Distributions
Figure 7.12 P(31.8 < x < 31.9)
The probability that x is between 31.8 and 31.9 is given by the area under the standard nor-
mal curve between z = —2.98 and z = —1.49. Thus, the required probability is
P(31.8 < x< 31.9) = P(-2.98 < z < -1.49)
= P(z < -1.49) - P(z < -2.98)
= .0681 - .0014 = .0667
Therefore, the probability is .0667 that the mean weight of a sample of 20 packages will be
between 31.8 and 31.9 ounces.
■ EXAMPLE 7-6
According to Sallie Mae surveys and Credit Bureau data, college students carried an average
of $3173 credit card debt in 2008. Suppose the probability distribution of the current credit
card debts of all college students in the United States is unknown but its mean is $3173 and
the standard deviation is $750. Let x be the mean credit card debt of a random sample of 400
U.S. college students.
(a) What is the probability that the mean of the current credit card debts for this sample
is within $70 of the population mean?
(b) What is the probability that the mean of the current credit card debts for this sample
is lower than the population mean by $50 or more?
Solution From the given information, for the current credit card debts of all college stu-
dents in the United States,
fjL = $3173 and a = $750
Although the shape of the probability distribution of the population (current credit card
debts of all college students in the United States) is unknown, the sampling distribution of x
is approximately normal because the sample is large (« > 30). Remember that when the sam-
ple is large, the central limit theorem applies. The mean and standard deviation of the sam-
pling distribution of x are, respectively,
jx x = jx = $3173 and
= $37.50
(a) The probability that the mean of the current credit card debts for this sample is within
$70 of the population mean is written as
P(3103 < x < 3243)
This probability is given by the area under the normal curve for x between
x = $3103 and x = $3243, as shown in Figure 7.13. We find this area as follows.
Calculating the probability
of x in an interval: n > 30.
7.5 Applications of the Sampling Distribution of x 319
$3103 ^ = $3173 $3243
I I I i
-1.87 1.87
Figure 7.13 P($3103 < x < $3243)
Hence, the required probability is
P($3103 < x < $3243) = P(-1.87 < z < 1.87)
= P(z < 1.87) - P(z < -1.87)
= .9693 - .0307 = .9386
Therefore, the probability that the mean of the current credit card debts for this sam-
ple is within $70 of the population mean is .9386.
(b) The probability that the mean of the current credit card debts for this sample is lower
than the population mean by $50 or more is written as
P(x < 3123)
This probability is given by the area under the normal curve for I to the left of
x = $3123, as shown in Figure 7.14. We find this area as follows:
$3123 p = $3173 x
-1 .33
Figure 7.14 P(x < $3123)
Hence, the required probability is
P(x < 3123) = P(z < -1.33) = .0918
Therefore, the probability that the mean of the current credit card debts for this sample
is lower than the population mean by $50 or more is .0918. H
320 Chapter 7 Sampling Distributions
EXERCISES
CONCEPTS AND PROCEDURES
7.40 If all possible samples of the same (large) size are selected from a population, what percentage of
all the sample means will be within 2.5 standard deviations of the population mean?
7.41 If all possible samples of the same (large) size are selected from a population, what percentage of
all the sample means will be within 1.5 standard deviations of the population mean?
7.42 For a population, N = 10,000, /jl = 124, and cr = 18. Find the z value for each of the following for
n = 36.
a. x = 128.60 b. x = 119.30 c. x = 116.88 d. x = 132.05
205,000, /jl = 66, and cr = 7. Find the z value for each of the following for
7.43 For a population, N
n = 49.
a. x = 68.44 b. x
58.75
62.35
d. x = 71.82
7.44 Let x be a continuous random variable that has a normal distribution with /x = 75 and cr = 14. As-
suming n/N £ .05, find the probability that the sample mean, x, for a random sample of 20 taken from
this population will be
a. between 68.5 and 77.3 b. less than 72.4
7.45 Let x be a continuous random variable that has a normal distribution with /x = 48 and cr = 8. As-
suming n/N £ .05, find the probability that the sample mean, x, for a random sample of 16 taken from
this population will be
a. between 49.6 and 52.2 b. more than 45.7
7.46 Let x be a continuous random variable that has a distribution skewed to the right with /jl = 60 and
cr — 10. Assuming n/N £ .05, find the probability that the sample mean, x, for a random sample of 40
taken from this population will be
a. less than 62.20 b. between 61.4 and 64.2
7.47 Let x be a continuous random variable that follows a distribution skewed to the left with /x = 90 and
cr = 18. Assuming n/N £ .05, find the probability that the sample mean, x, for a random sample of 64
taken from this population will be
a. less than 82.3 b. greater than 86.7
■ APPLICATIONS
7.48 According to the article by Laroche et al. mentioned in Exercise 7.37, the average daily fat intake
of U.S. adults with children in the household is 91.4 grams, with a standard deviation of 93.25 grams.
Find the probability that the average daily fat intake of a random sample of 75 U.S. adults with children
in the household is
a. less than 80 grams b. more than 100 grams c. 95 to 102 grams
7.49 The GPAs of all students enrolled at a large university have an approximately normal distribution
with a mean of 3.02 and a standard deviation of .29. Find the probability that the mean GPA of a random
sample of 20 students selected from this university is
a. 3.10 or higher b. 2.90 or lower c. 2.95 to 3.11
7.50 The delivery times for all food orders at a fast-food restaurant during the lunch hour are normally
distributed with a mean of 7.7 minutes and a standard deviation of 2.1 minutes. Find the probability that
the mean delivery time for a random sample of 16 such orders at this restaurant is
a. between 7 and 8 minutes
b. within 1 minute of the population mean
c. less than the population mean by 1 minute or more
7.51 As mentioned in Exercise 7.22, the average cost of going to a minor league baseball game for a fam-
ily of four was $55 in 2009. Suppose that the standard deviation of such costs is $13.25. Find the probabil-
ity that the average cost of going to a minor league baseball game for 33 randomly selected such families is
a. more than $60 b. less than $52 c. $54 to $57.99
7.52 The times that college students spend studying per week have a distribution that is skewed to the
right with a mean of 8.4 hours and a standard deviation of 2.7 hours. Find the probability that the mean
time spent studying per week for a random sample of 45 students would be
a. between 8 and 9 hours b. less than 8 hours
7.6 Population and Sample Proportions
321
7.53 The credit card debts of all college students have a distribution that is skewed to the right with a
mean of $2840 and a standard deviation of $672. Find the probability that the mean credit card debt for
a random sample of 36 college students would be
a. between $2600 and $2950
b. less than $3060
7.54 As mentioned in Exercise 7.39, the annual per capita (average per person) chewing gum consump-
tion in the United States is 200 pieces. Suppose that the standard deviation of per capita consumption of
chewing gum is 145 pieces per year. Find the probability that the average annual chewing gum consump-
tion of 84 randomly selected Americans is
a. 160 to 170 pieces
b. more than 220 pieces
c. at most 150 pieces
7.55 The amounts of electricity bills for all households in a city have a skewed probability distribution
with a mean of $140 and a standard deviation of $30. Find the probability that the mean amount of elec-
tric bills for a random sample of 75 households selected from this city will be
a. between $132 and $136
b. within $6 of the population mean
c. more than the population mean by at least $4
7.56 As mentioned in Exercise 7.19, the average retail price of a gallon of whole milk in the United
States was $3,084 in April 2009. Suppose that the current distribution of the retail prices of a gallon
of whole milk in the United States has a mean of $3,084 and a standard deviation of $.263. Find
the probability that the average retail price of a gallon of whole milk from a random sample of 47
stores is
a. less than $3.00
b. more than $3.20
c. $3.10 to $3.15
7.57 As mentioned in Exercise 7.33, among college students who hold part-time jobs during the school
year, the distribution of the time spent working per week is approximately normally distributed with a
mean of 20.20 hours and a standard deviation of 2.6 hours. Find the probability that the average time
spent working per week for 18 randomly selected college students who hold part-time jobs during the
school year is
a. not within 1 hour of the population mean
b. 20.0 to 20.5 hours
c. at least 22 hours
d. no more than 21 hours
7.58 Johnson Electronics Corporation makes electric tubes. It is known that the standard deviation of the
lives of these tubes is 150 hours. The company's research department takes a sample of 100 such tubes
and finds that the mean life of these tubes is 2250 hours. What is the probability that this sample mean is
within 25 hours of the mean life of all tubes produced by this company?
7.59 A machine at Katz Steel Corporation makes 3-inch-long nails. The probability distribution of the
lengths of these nails is normal with a mean of 3 inches and a standard deviation of . 1 inch. The quality
control inspector takes a sample of 25 nails once a week and calculates the mean length of these nails. If
the mean of this sample is either less than 2.95 inches or greater than 3.05 inches, the inspector concludes
that the machine needs an adjustment. What is the probability that based on a sample of 25 nails, the in-
spector will conclude that the machine needs an adjustment?
7.6 Population and Sample Proportions
The concept of proportion is the same as the concept of relative frequency discussed in
Chapter 2 and the concept of probability of success in a binomial experiment. The rela-
tive frequency of a category or class gives the proportion of the sample or population that
belongs to that category or class. Similarly, the probability of success in a binomial
experiment represents the proportion of the sample or population that possesses a given
characteristic.
322 Chapter 7 Sampling Distributions
The population proportion, denoted by p, is obtained by taking the ratio of the number
of elements in a population with a specific characteristic to the total number of elements in
the population. The sample proportion, denoted by p (pronounced p hat), gives a similar
ratio for a sample.
Population and Sample Proportions The population and sample proportions, denoted by p
and p, respectively, are calculated as
X
— and
N
where
N = total number of elements in the population
n = total number of elements in the sample
X = number of elements in the population that possess a specific characteristic
x = number of elements in the sample that possess a specific characteristic
Example 7-7 illustrates the calculation of the population and sample proportions.
■ EXAMPLE 7-7
Suppose a total of 789,654 families live in a particular city and 563,282 of them own homes.
Calculating the population
. , A sample of 240 families is selected from this city, and 158 of them own homes. Find the
ana sample proportions. r J
proportion of families who own homes in the population and in the sample.
Solution For the population of this city,
N = population size = 789,654
X = families in the population who own homes = 563,282
The proportion of all families in this city who own homes is
X 563,282
p = — = = .71
' N 789,654
Now, a sample of 240 families is taken from this city, and 158 of them are home-owners.
Then,
n = sample size = 240
x = families in the sample who own homes = 158
The sample proportion is
F n 240
As in the case of the mean, the difference between the sample proportion and the
corresponding population proportion gives the sampling error, assuming that the sample is ran-
dom and no nonsampling error has been made. That is, in the case of the proportion,
Sampling error = p — p
For instance, for Example 7-7,
Sampling error = p — p = .66 — .71 = —.05
7.7 Mean, Standard Deviation, and Shape of the Sampling Distribution of p 323
7.7 Mean, Standard Deviation, and Shape
of the Sampling Distribution of p
This section discusses the sampling distribution of the sample proportion and the mean, stan-
dard deviation, and shape of this sampling distribution.
7.7.1 Sampling Distribution of p
Just like the sample mean x, the sample proportion p is a random variable. Hence, it possesses
a probability distribution, which is called its sampling distribution.
Definition
Sampling Distribution of the Sample Proportion, p The probability distribution of the sample
proportion, p, is called its sampling distribution. It gives the various values that p can assume
and their probabilities.
The value of p calculated for a particular sample depends on what elements of the popu-
lation are included in that sample. Example 7-8 illustrates the concept of the sampling
distribution of p.
■ EXAMPLE 7-8
Boe Consultant Associates has five employees. Table 7.6 gives the names of these five
employees and information concerning their knowledge of statistics.
Table 7.6 Information on the
Five Employees of
Boe Consultant
Associates
Name
Knows Statistics
Ally
Yes
John
No
Susan
No
Lee
Yes
Tom
Yes
If we define the population proportion, p, as the proportion of employees who know
statistics, then
p = 3/5 = .60
Now, suppose we draw all possible samples of three employees each and compute the pro-
portion of employees, for each sample, who know statistics. The total number of samples of
size three that can be drawn from the population of five employees is
5! 5 • 4 • 3 • 2 • 1
Total number of samples = <Ci = — ; — = =10
F 5 3 3!(5 - 3)! 3 • 2 • 1 • 2 • 1
Table 7.7 lists these 10 possible samples and the proportion of employees who know
statistics for each of those samples. Note that we have rounded the values of p to two dec-
imal places.
Illustrating the
sampling distribution of p.
324 Chapter 7 Sampling Distributions
Table 7.7 All Possible Samples of Size 3 and
the Value of p for Each Sample
Proportion Who
Know Statistics
Sample p
Ally, John, Susan
1/3 =
.33
Ally, John, Lee
2/3 =
.67
Ally, John, Tom
2/3 =
.67
Ally, Susan, Lee
2/3 =
.67
Ally, Susan, Tom
2/3 =
.67
Ally, Lee, Tom
3/3 =
1.00
John, Susan, Lee
1/3 =
.33
John, Susan, Tom
1/3 =
.33
John, Lee, Tom
2/3 =
.67
Susan, Lee, Tom
2/3 =
.67
Using Table 7.7, we prepare the frequency distribution of p as recorded in Table 7.8, along
with the relative frequencies of classes, which are obtained by dividing the frequencies of
classes by the population size. The relative frequencies are used as probabilities and listed in
Table 7.9. This table gives the sampling distribution of p.
Table 7.8
Frequency and Relative
Frequency Distributions of p
When the Sample Size Is 3
Table 7.9
Sampling Distri-
bution of p When
the Sample Size
Relative
Is 3
P
f
Frequency
P
Pip)
.33
3
3/10= .30
.33
.30
.67
6
6/10 = .60
.67
.60
1.00
1
1/10= .10
1.00
.10
2/= 10
Sum = 1.00
%P(p) = 1.00
7.7.2 Mean and Standard Deviation of p
The mean of/?, which is the same as the mean of the sampling distribution of p, is always equal
to the population proportion, p, just as the mean of the sampling distribution of x is always
equal to the population mean, fi.
Mean of the Sample Proportion The mean of the sample proportion, p, is denoted by i±p and is
equal to the population proportion, p. Thus,
Hp = P
The sample proportion, p, is called an estimator of the population proportion, p. As men-
tioned earlier, when the expected value (or mean) of a sample statistic is equal to the value of
the corresponding population parameter, that sample statistic is said to be an unbiased estima-
tor. Since for the sample proportion /i- = p, p is an unbiased estimator of p.
The standard deviation of p, denoted by cr~, is given by the following formula. This formula
is true only when the sample size is small compared to the population size. As we know from Sec-
tion 7.3, the sample size is said to be small compared to the population size if n/N s .05.
7.7 Mean, Standard Deviation, and Shape of the Sampling Distribution of p
Standard Deviation of the Sample Proportion The standard deviation of the sample proportion,
p, is denoted by cr» and is given by the formula
where p is the population proportion, q = 1 — p, and n is the sample size. This formula is used
when n/N s .05, where N is the population size.
However, if n/N is greater than .05, then cr- is calculated as follows:
I N - n
Vjv- 1
where the factor
j N - n
is called the finite-population correction factor.
In almost all cases, the sample size is small compared to the population size and, conse-
quently, the formula used to calculate cr~ is \fpqfn.
As mentioned earlier, if the standard deviation of a sample statistic decreases as the sam-
ple size is increased, that statistic is said to be a consistent estimator. It is obvious from the
above formula for <x- that as n increases, the value of ypqjn decreases. Thus, the sample pro-
portion, p, is a consistent estimator of the population proportion, p.
7.7.3 Shape of the Sampling Distribution of p
The shape of the sampling distribution of p is inferred from the central limit theorem.
Central Limit Theorem for Sample Proportion According to the central limit theorem, the sam-
pling distribution of p is approximately normal for a sufficiently large sample size. In the case
of proportion, the sample size is considered to be sufficiently large if np and nq are both greater
than 5 — that is, if
np > 5 and nq > 5
Note that the sampling distribution of p will be approximately normal if np > 5 and nq > 5.
This is the same condition that was required for the application of the normal approximation to
the binomial probability distribution in Chapter 6.
Example 7-9 shows the calculation of the mean and standard deviation of p and describes
the shape of its sampling distribution.
■ EXAMPLE 7-9
According to a survey by Harris Interactive conducted in February 2009 for the charitable
agency World Vision, 56% of U.S. teens volunteer time for charitable causes. Assume that this
result is true for the current population of U.S. teens. Let p be the proportion of U.S. teens in
a random sample of 1500 who volunteer time for charitable causes. Find the mean and stan-
dard deviation of p, and describe the shape of its sampling distribution.
Finding the mean,
standard deviation, and
shape of the sampling
distribution of p.
Solution
Then,
Let p be the proportion of all U.S. teens who volunteer time for charitable causes.
p = .56 and q = 1 - p = 1 - .56 = .44
326 Chapter 7 Sampling Distributions
The mean of the sampling distribution of p is
fi f = p = .56
The standard deviation of p is
= .0128
V n V
The values of tip and nq are
np = 1500(.56) = 840 and nq = 1500(.44) = 660
Because np and nq are both greater than 5, we can apply the central limit theorem to make an
inference about the shape of the sampling distribution of p. Therefore, the sampling distribu-
tion of p is approximately normal with a mean of .56 and a standard deviation of .0128, as
shown in Figure 7.15.
EXERCISES
CONCEPTS AND PROCEDURES
7.60 In a population of 1000 subjects, 640 possess a certain characteristic. A sample of 40 subjects se-
lected from this population has 24 subjects who possess the same characteristic. What are the values of
the population and sample proportions?
7.61 In a population of 5000 subjects, 600 possess a certain characteristic. A sample of 120 subjects se-
lected from this population contains 18 subjects who possess the same characteristic. What are the values
of the population and sample proportions?
7.62 In a population of 18,700 subjects, 30% possess a certain characteristic. In a sample of 250 subjects
selected from this population, 25% possess the same characteristic. How many subjects in the population
and sample, respectively, possess this characteristic?
7.63 In a population of 9500 subjects, 75% possess a certain characteristic. In a sample of 400 subjects
selected from this population, 78% possess the same characteristic. How many subjects in the population
and sample, respectively, possess this characteristic?
7.64 Let p be the proportion of elements in a sample that possess a characteristic.
a. What is the mean of pi
b. What is the standard deviation of pi Assume n/N ^ .05.
c. What condition(s) must hold true for the sampling distribution of p to be approximately normal?
7.65 For a population, N = 12,000 and p = .71. A random sample of 900 elements selected from this
population gave p = .66. Find the sampling error.
7.66 For a population, N = 2800 and p = .29. A random sample of 80 elements selected from this pop-
ulation gave p = .33. Find the sampling error.
7.67 What is the estimator of the population proportion? Is this estimator an unbiased estimator of pi Ex-
plain why or why not.
7.68 Is the sample proportion a consistent estimator of the population proportion? Explain why or why not.
7.69 How does the value of crj change as the sample size increases? Explain. Assume n/N S .05.
7.7 Mean, Standard Deviation, and Shape of the Sampling Distribution of p 327
7.70 Consider a large population with p = .63. Assuming n/N S .05, find the mean and standard devia-
tion of the sample proportion p for a sample size of
a. 100 b. 900
7.71 Consider a large population with p = .21. Assuming n/N S .05, find the mean and standard devia-
tion of the sample proportion p for a sample size of
a. 400 b. 750
7.72 A population of N = 4000 has a population proportion equal to .12. In each of the following cases,
which formula will you use to calculate ap and why? Using the appropriate formula, calculate o-f, for each
of these cases.
a. n = 800 b. n = 30
7.73 A population of N = 1400 has a population proportion equal to .47. In each of the following cases,
which formula will you use to calculate cr /5 and why? Using the appropriate formula, calculate o-f, for each
of these cases.
a. n = 90 b. n = 50
7.74 According to the central limit theorem, the sampling distribution of p is approximately normal
when the sample is large. What is considered a large sample in the case of the proportion? Briefly
explain.
7.75 Indicate in which of the following cases the central limit theorem will apply to describe the sam-
pling distribution of the sample proportion.
a. n = 400 and p = .28 b. n = 80 and p = .05
c. n = 60 and p = .12 d. n = 100 and p = .035
7.76 Indicate in which of the following cases the central limit theorem will apply to describe the sam-
pling distribution of the sample proportion.
a. n = 20 and p = .45 b. n = 75 and p = .22
c. n = 350 and p = .01 d. n = 200 and p = .022
■ APPLICATIONS
7.77 A company manufactured six television sets on a given day, and these TV sets were inspected for
being good or defective. The results of the inspection follow.
Good Good Defective Defective Good Good
a. What proportion of these TV sets are good?
b. How many total samples (without replacement) of size five can be selected from this population?
c. List all the possible samples of size five that can be selected from this population and calculate
the sample proportion, p, of television sets that are good for each sample. Prepare the sampling
distribution of p.
d. For each sample listed in part c, calculate the sampling error.
7.78 Investigation of all five major fires in a western desert during one of the recent summers found the
following causes.
Arson Accident Accident Arson Accident
a. What proportion of those fires were due to arson?
b. How many total samples (without replacement) of size three can be selected from this population?
c. List all the possible samples of size three that can be selected from this population and calculate
the sample proportion p of the fires due to arson for each sample. Prepare the table that gives the
sampling distribution of p.
d. For each sample listed in part c, calculate the sampling error.
7.79 According to a 2008 survey by the Royal Society of Chemistry, 30% of adults in Great Britain
said that they typically run the water for a period of 6 to 10 minutes while they take a shower
(http://www.rsc.org/AboutUs/News/PressReleases/2008/EuropeanShowerHabits.asp). Assume that this
percentage is true for the current population of adults in Great Britain. Let p be the proportion in a ran-
dom sample of 180 adults from Great Britain who typically run the water for a period of 6 to 10 minutes
while they take a shower. Find the mean and standard deviation of the sampling distribution of p and de-
scribe its shape.
7.80 In an observational study at Turner Field in Atlanta, Georgia, 43% of the men were observed not
washing their hands after going to the bathroom (Source: Harris Interactive). Assume that this percentage
328 Chapter 7 Sampling Distributions
is true for the current population of U.S. men. Let p be the proportion in a random sample of 1 10 U.S.
men who do not wash their hands after going to the bathroom. Find the mean and standard deviation of
the sampling distribution of p, and describe its shape.
7.81 A 2009 nonscientific poll on the Web site of the Daily Gazette of Schenectady, New York, asked
readers the following question: "Are you less inclined to buy a General Motors or Chrysler vehicle now
that they have filed for bankruptcy?" Of the readers who responded, 56.1% answered Yes (http://www.
dailygazette.com/polls/2009/jun/Bankruptcy/). Assume that this result is true for the current population
of vehicle owners in the United States. Let p be the proportion in a random sample of 340 U.S. vehi-
cle owners who are less inclined to buy a General Motors or Chrysler vehicle after these corporations
filed for bankruptcy. Find the mean and standard deviation of the sampling distribution of p, and de-
scribe its shape.
7.82 According to the American Diabetes Association (www.diabetes.org), 23.1% of Americans aged
60 years or older had diabetes in 2007. Assume that this percentage is true for the current population of
Americans aged 60 years or older. Let p be the proportion in a random sample of 460 Americans aged
60 years or older who have diabetes. Find the mean and standard deviation of the sampling distribution
of p, and describe its shape.
7.8 Applications of the Sampling
Distribution of p
As mentioned in Section 7.5, when we conduct a study, we usually take only one sample and
make all decisions or inferences on the basis of the results of that one sample. We use the con-
cepts of the mean, standard deviation, and shape of the sampling distribution of p to determine
the probability that the value of p computed from one sample falls within a given interval. Ex-
amples 7-10 and 7-11 illustrate this application.
Calculating the probability
that p is in an interval.
■ EXAMPLE 7-10
According to the BBMG Conscious Consumer Report, 5 1 % of the adults surveyed said that
they are willing to pay more for products with social and environmental benefits despite the
current tough economic times (USA TODAY, June 8, 2009). Suppose this result is true for the
current population of adult Americans. Let p be the proportion in a random sample of 1050
adult Americans who will hold the said opinion. Find the probability that the value of p is be-
tween .53 and .55.
Solution From the given information,
n = 1050, p = .51, and
9
1
1 - .51
.49
where p is the proportion of all adult Americans who will hold the said opinion.
The mean of the sample proportion p is
The standard deviation of p is
The values of np and nq are
.51
(■51)049)
.01542725
np = 1050 (.51) = 535.5 and nq = 1050 (.49) = 514.5
Because np and nq are both greater than 5, we can infer from the central limit theorem that
the sampling distribution of p is approximately normal. The probability that p is between .53
and .55 is given by the area under the normal curve for p between p = .53 and p = .55, as
shown in Figure 7.16.
7.8 Applications of the Sampling Distribution of p 329
Figure 7.16 P(.53 < p < .55)
Approximately *Y \
.01542725
normal distribution /
1
y ff ,5 =
J
.51 — 1
.53 .55
The first step in finding the area under the normal curve between /> = .53 and p = .55 is
to convert these two values to their respective z values. The z value for p is computed using
the following formula.
z Value for a Value of p The z value for a value of p is calculated as
p-p
z =
o-f,
The two values of p are converted to their respective z values, and then the area under the nor-
mal curve between these two points is found using the normal distribution table.
.53 - .51
Forp = .53: z= = 1.30
For/3 = .55:
.01542725
.55 - .51
.01542725
2.59
Thus, the probability that p is between .53 and .55 is given by the area under the standard
normal curve between z = 1.30 and z = 2.59. This area is shown in Figure 7.17. The required
probability is
P(.53 < p < .55) = P(1.30 < z < 2.59) = P{z < 2.59) - P{z < 1.30)
= .9952 - .9032 = .0920
i — Shaded area
is .0920
Figure 7.17 P(.53 < p < .55)
— I 1 —
1.30 2.59
Thus, the probability is .0920 that the proportion of U.S. adults in a random sample of
1050 who will be willing to pay more for products with social and environmental benefits de-
spite the current tough economic times is between .53 and .55. I
■ EXAMPLE 7-11
Maureen Webster, who is running for mayor in a large city, claims that she is favored by 53%
of all eligible voters of that city. Assume that this claim is true. What is the probability that
in a random sample of 400 registered voters taken from this city, less than 49% will favor
Maureen Webster?
Calculating the probability that
p is less than a certain value.
330 Chapter 7 Sampling Distributions
Solution Let p be the proportion of all eligible voters who favor Maureen Webster. Then,
p = .53 and q=l-p = l- .53 = .47
The mean of the sampling distribution of the sample proportion p is
fi0 = p = .53
The population of all voters is large (because the city is large), and the sample size is small
compared to the population. Consequently, we can assume that n/N ^ .05. Hence, the stan-
dard deviation of p is calculated as
a /C53V.471
= .02495496
From the central limit theorem, the shape of the sampling distribution of p is approximately
normal. The probability that p is less than .49 is given by the area under the normal distribu-
tion curve for p to the left of p = .49, as shown in Figure 7.18. The z value for p = .49 is
P ~ p .49 - .53
z = = = —1.60
o> .02495496
Thus, the required probability from table IV is
P(p < .49) = P(z < -1.60)
= .0548
Hence, the probability that less than 49% of the voters in a random sample of 400 will
favor Maureen Webster is .0548.
EXERCISES
CONCEPTS AND PROCEDURES
7.83 If all possible samples of the same (large) size are selected from a population, what percentage of
all sample proportions will be within 2.0 standard deviations of the population proportion?
7.84 If all possible samples of the same (large) size are selected from a population, what percentage of
all sample proportions will be within 3.0 standard deviations of the population proportion?
7.85 For a population, N = 30,000 and p = .59. Find the z value for each of the following for n = 100.
a. p = .56 b. p = .68 c. p = .53 d. p = .65
7.86 For a population, N = 18,000 and p = .25. Find the z value for each of the following for n = 70.
a. p = .26 b. p = .32 c. p = .17 d. p = .20
■ APPLICATIONS
7.87 As mentioned in Exercise 7.79, 30% of adults in Great Britain stated that they typically run the wa-
ter for a period of 6 to 10 minutes before they take a shower. Let p be the proportion in a random sample
Uses and Misuses 331
of 180 adults from Great Britain who typically run the water for a period of 6 to 10 minutes before they
take a shower. Find the probability that the value of p will be
a. greater than .35 b. between .22 and .27
7.88 A survey of all medium- and large-sized corporations showed that 64% of them offer retirement plans
to their employees. Let p be the proportion in a random sample of 50 such corporations that offer retire-
ment plans to their employees. Find the probability that the value of p will be
a. between .54 and .61 b. greater than .71
7.89 As mentioned in Exercise 7.80, in an observational study at Turner Field in Atlanta, Georgia, 43%
of the men were observed not washing their hands after going to the bathroom. Assume that the percent-
age of all U.S. men who do not wash their hands after going to the bathroom is 43%. Let p be the pro-
portion in a random sample of 110 U.S. men who do not wash their hands after going to the bathroom.
Find the probability that the value of p will be
a. less than .30 b. between .45 and .50
7.90 Dartmouth Distribution Warehouse makes deliveries of a large number of products to its customers.
It is known that 85% of all the orders it receives from its customers are delivered on time. Let p be the
proportion of orders in a random sample of 100 that are delivered on time. Find the probability that the
value of p will be
a. between .81 and .88 b. less than .87
7.91 Brooklyn Corporation manufactures CDs. The machine that is used to make these CDs is known to
produce 6% defective CDs. The quality control inspector selects a sample of 100 CDs every week and in-
spects them for being good or defective. If 8% or more of the CDs in the sample are defective, the process
is stopped and the machine is readjusted. What is the probability that based on a sample of 100 CDs, the
process will be stopped to readjust the machine?
7.92 Mong Corporation makes auto batteries. The company claims that 80% of its LL70 batteries are good
for 70 months or longer. Assume that this claim is true. Let p be the proportion in a sample of 100 such
batteries that are good for 70 months or longer.
a. What is the probability that this sample proportion is within .05 of the population proportion?
b. What is the probability that this sample proportion is less than the population proportion by .06
or more?
c. What is the probability that this sample proportion is greater than the population proportion by
.07 or more?
USES AND MISUSES... BEWARE OF BIAS
Mathematics tells us that the sample mean, x, is an unbiased and
consistent estimator for the population mean, fj.. This is great news
because it allows us to estimate properties of a population based on
those of a sample; this is the essence of statistics. But statistics al-
ways makes a number of assumptions about the sample from which
the mean and standard deviation are calculated. Failure to respect
these assumptions can introduce bias in your calculations. In statis-
tics, bias means a deviation of the expected value of a statistical es-
timator from the parameter it estimates.
Let's say you are a quality control manager for a refrigerator
parts company. One of the parts that you manufacture has a spec-
ification that the length of the part be 2.0 centimeters plus or
minus .025 centimeter. The manufacturer expects that the parts it
receives have a mean length of 2.0 centimeters and a small variation
around that mean. The manufacturing process is to mold the part
to something a little bit bigger than necessary— say, 2.1 centimeters—
and finish the process by hand. Because the action of cutting
material is irreversible, the machinists tend to miss their target by
approximately .01 centimeter, so the mean length of the parts is
not 2.0 centimeters, but rather 2.01 centimeters. It is your job to
catch this.
One of your quality control procedures is to select completed
parts randomly and test them against specification. Unfortunately,
your measurement device is also subject to variation and might con-
sistently underestimate the length of the parts. If your measurements
are consistently .01 centimeter too short, your sample mean will not
catch the manufacturing error in the population of parts.
The solution to the manufacturing problem is relatively straight-
forward: Be certain to calibrate your measurement instrument. Cali-
bration becomes very difficult when working with people. It is known
that people tend to overestimate the number of times that they vote
and underestimate the time it takes to complete a project. Basing
statistical results on this type of data can result in distorted estimates
of the properties of your population. It is very important to be care-
ful to weed out bias in your data because once it gets into your cal-
culations, it is very hard to get it out.
332 Chapter 7 Sampling Distributions
Glossary
Central limit theorem The theorem from which it is inferred that
for a large sample size (n £ 30), the shape of the sampling distri-
bution of x is approximately normal. Also, by the same theorem, the
shape of the sampling distribution of p is approximately normal for
a sample for which np > 5 and nq > 5.
Consistent estimator A sample statistic with a standard deviation
that decreases as the sample size increases.
Estimator The sample statistic that is used to estimate a popula-
tion parameter.
Mean of p The mean of the sampling distribution of p, denoted
by (U, /5 , is equal to the population proportion p.
Mean of x The mean of the sampling distribution of x, denoted by
is equal to the population mean p..
Nonsampling errors The errors that occur during the collection,
recording, and tabulation of data.
Population distribution The probability distribution of the popu-
lation data.
Population proportion p The ratio of the number of elements in
a population with a specific characteristic to the total number of el-
ements in the population.
Sample proportion p The ratio of the number of elements in a
sample with a specific characteristic to the total number of elements
in that sample.
Sampling distribution of p The probability distribution of all the
values of p calculated from all possible samples of the same size se-
lected from a population.
Sampling distribution of x The probability distribution of all the
values of x calculated from all possible samples of the same size se-
lected from a population.
Sampling error The difference between the value of a sample sta-
tistic calculated from a random sample and the value of the corre-
sponding population parameter. This type of error occurs due to
chance.
Standard deviation of p The standard deviation of the sampling
distribution of p, denoted by crp, is equal to \fpqfn when n/N ^
.05.
Standard deviation of x The standard deviation of the sampling
distribution of x, denoted by a 7r , is equal to cr/Vn when n/N S .05.
Unbiased estimator An estimator with an expected value (or
mean) that is equal to the value of the corresponding population pa-
rameter.
Supplementary Exercises
7.93 The print on the package of 100-watt General Electric soft-white lightbulbs claims that these
bulbs have an average life of 750 hours. Assume that the lives of all such bulbs have a normal distri-
bution with a mean of 750 hours and a standard deviation of 55 hours. Let x be the mean life of a ran-
dom sample of 25 such bulbs. Find the mean and standard deviation of x, and describe the shape of
its sampling distribution.
7.94 According to a 2004 survey by the telecommunications division of British Gas (Source:
http://www.literacytrust.org.Uk/Database/texting.html#quarter), Britons spend an average of 225 minutes
per day communicating electronically (on a landline phone, on a mobile phone, by emailing, or by texting).
Assume that currently such communication times for all Britons are normally distributed with a mean of
225 minutes per day and a standard deviation of 62 minutes per day. Let x be the average time spent per
day communicating electronically by 20 randomly selected Britons. Find the mean and the standard devi-
ation of the sampling distribution of x. What is the shape of the sampling distribution of x?
7.95 Refer to Exercise 7.93. The print on the package of 100-watt General Electric soft-white light-bulbs
says that these bulbs have an average life of 750 hours. Assume that the lives of all such bulbs have a nor-
mal distribution with a mean of 750 hours and a standard deviation of 55 hours. Find the probability that
the mean life of a random sample of 25 such bulbs will be
a. greater than 735 hours
b. between 725 and 740 hours
c. within 15 hours of the population mean
d. less than the population mean by 20 hours or more
7.96 Refer to Exercise 7.94. On average, Britons spend 225 minutes per day communicating electroni-
cally. Assume that currently such communication times for all Britons are normally distributed with a
mean of 225 minutes per day and a standard deviation of 62 minutes per day. Find the probability that the
mean time spent communicating electronically per day by a random sample of 20 Britons will be
a. less than 200 minutes
b. between 230 and 240 minutes
c. within 20 minutes of the population mean
d. more than 260 minutes
Supplementary Exercises 333
7.97 According to an article on www.PCMag.com, Facebook users spend an average of 190 minutes
per month checking and updating their Facebook pages (Source: http://www.pcmag.com/article2/
0,2817, 2342757,00.asp). Suppose that the current distribution of times spent per month checking and
updating their Facebook pages by all users is normally distributed with a mean of 190 minutes and a stan-
dard deviation of 53.4 minutes. Find the probability that the mean time spent per month checking and
updating their Facebook pages by a random sample of 12 Facebook users will be
a. within 10 minutes of the population mean
b. more than 240 minutes
c. at least 20 minutes different than the population mean
d. less than 207 minutes
7.98 A machine at Keats Corporation fills 64-ounce detergent jugs. The probability distribution of the
amount of detergent in these jugs is normal with a mean of 64 ounces and a standard deviation of
.4 ounce. The quality control inspector takes a sample of 16 jugs once a week and measures the amount
of detergent in these jugs. If the mean of this sample is either less than 63.75 ounces or greater than
64.25 ounces, the inspector concludes that the machine needs an adjustment. What is the probability
that based on a sample of 16 jugs, the inspector will conclude that the machine needs an adjustment
when actually it does not?
7.99 Suppose that 88% of the cases of car burglar alarms that go off are false. Let p be the proportion of
false alarms in a random sample of 80 cases of car burglar alarms that go off. Calculate the mean and
standard deviation of p, and describe the shape of its sampling distribution.
7.100 Seventy percent of adults favor some kind of government control on the prices of medicines. As-
sume that this percentage is true for the current population of all adults. Let p be the proportion of adults
in a random sample of 400 who favor government control on the prices of medicines. Calculate the mean
and standard deviation of p and describe the shape of its sampling distribution.
7.101 Refer to Exercise 7.100. Seventy percent of adults favor some kind of government control
on the prices of medicines. Assume that this percentage is true for the current population of all
adults.
a. Find the probability that the proportion of adults in a random sample of 400 who favor some
kind of government control on the prices of medicines is
i. less than .65 ii. between .73 and .76
b. What is the probability that the proportion of adults in a random sample of 400 who favor
some kind of government control is within .06 of the population proportion?
c. What is the probability that the sample proportion is greater than the population proportion by
.05 or more? Assume that sample includes 400 adults.
7.102 Gluten sensitivity, which is also known as wheat intolerance, affects approximately 15% of people
in the United States (Source: http://www.foodintol.com/wheat.asp). Let p be the proportion in a random
sample of 800 individuals who have gluten sensitivity. Find the probability that the value of p is
a. within .02 of the population proportion
b. not within .02 of the population proportion
c. greater than the population proportion by .025 or more
d. less than the population proportion by .03 or more
Advanced Exercises
7.103 Let jx be the mean annual salary of Major League Baseball players for 2009. Assume that the stan-
dard deviation of the salaries of these players is $2,845,000. What is the probability that the 2009 mean
salary of a random sample of 26 baseball players was within $500,000 of the population mean, pi As-
sume that n/N — .05.
7.104 The test scores for 300 students were entered into a computer, analyzed, and stored in a file. Un-
fortunately, someone accidentally erased a major portion of this file from the computer. The only infor-
mation that is available is that 30% of the scores were below 65 and 15% of the scores were above 90.
Assuming the scores are normally distributed, find their mean and standard deviation.
7.105 A chemist has a 10-gallon sample of river water taken just downstream from the outflow of a chem-
ical plant. He is concerned about the concentration, c (in parts per million), of a certain toxic substance
in the water. He wants to take several measurements, find the mean concentration of the toxic substance
for this sample, and have a 95% chance of being within .5 part per million of the true mean value of c. If
the concentration of the toxic substance in all measurements is normally distributed with <x = .8 part per
million, how many measurements are necessary to achieve this goal?
334
Chapter 7 Sampling Distributions
7.106 A television reporter is covering the election for mayor of a large city and will conduct an exit poll
(interviews with voters immediately after they vote) to make an early prediction of the outcome. Assume
that the eventual winner of the election will get 60% of the votes.
a. What is the probability that a prediction based on an exit poll of a random sample of 25 vot-
ers will be correct? In other words, what is the probability that 13 or more of the 25 voters in
the sample will have voted for the eventual winner?
b. How large a sample would the reporter have to take so that the probability of correctly pre-
dicting the outcome would be .95 or higher?
7.107 A city is planning to build a hydroelectric power plant. A local newspaper found that 53% of the
voters in this city favor the construction of this plant. Assume that this result holds true for the popula-
tion of all voters in this city.
a. What is the probability that more than 50% of the voters in a random sample of 200 voters
selected from this city will favor the construction of this plant?
b. A politician would like to take a random sample of voters in which more than 50% would fa-
vor the plant construction. How large a sample should be selected so that the politician is 95%
sure of this outcome?
7.108 Refer to Exercise 6.92. Otto is trying out for the javelin throw to compete in the Olympics. The
lengths of his javelin throws are normally distributed with a mean of 290 feet and a standard devia-
tion of 10 feet. What is the probability that the total length of three of his throws will exceed 885 feet?
7.109 A certain elevator has a maximum legal carrying capacity of 6000 pounds. Suppose that the popu-
lation of all people who ride this elevator have a mean weight of 160 pounds with a standard deviation of
25 pounds. If 35 of these people board the elevator, what is the probability that their combined weight will
exceed 6000 pounds? Assume that the 35 people constitute a random sample from the population.
7.110 According to a National Center for Education Statistics survey released in 2007, 41.5% of Utah
households used a public library or bookmobile within the past 1 month (http://harvester.census.gov/imls/
pubs/Publications/2007327.pdf). Suppose that this percentage is true for the current population of house-
holds in Utah.
a. Suppose that 5 1 .4% in a sample of 70 Utah households have used a public library or bookmo-
bile within the past 1 month. How likely is it for the sample proportion in a sample of 70 to
be .514 or more when the population proportion is .415?
b. Refer to part a. How likely is it for the sample proportion to be .514 or more when the sample
size is 250 and the population proportion is .415?
c. What is the smallest sample size that will produce a sample proportion of .514 or more in no
more than 1% of all sample surveys of that size?
7.111 Refer to the sampling distribution discussed in Section 7.1. Calculate and replace the sample means
in Table 7.3 with the sample medians, and then calculate the average of these sample medians. Does this
average of the medians equal the population mean? If yes, why does this make sense? If no, how could
you change exactly two of the five data values in this example so that the average of the sample medians
equals the population mean?
7.112 Suppose you want to calculate P(a £ x < b), where a and b are two numbers and x has a distribu-
tion with mean /x and standard deviation cr. If a < /jl < b (i.e., yii lies in the interval a to b), what hap-
pens to the probability P(a < x < b) as the sample size becomes larger?
Self-Review Test
1. A sampling distribution is the probability distribution of
a. a population parameter b. a sample statistic c. any random variable
2. Nonsampling errors are
a. the errors that occur because the sample size is too large in relation to the population size
b. the errors made while collecting, recording, and tabulating data
c. the errors that occur because an untrained person conducts the survey
3. A sampling error is
a. the difference between the value of a sample statistic based on a random sample and the value of
the corresponding population parameter
b. the error made while collecting, recording, and tabulating data
c. the error that occurs because the sample is too small
Self-Review Test 335
4. The mean of the sampling distribution of x is always equal to
a. jjl b. p — 5 c. cr/Vn
5. The condition for the standard deviation of the sample mean to be cr/Vn is that
a. np > 5 b. n/N < .05 c. n > 30
6. The standard deviation of the sampling distribution of the sample mean decreases when
a. x increases b. n increases c. n decreases
7. When samples are selected from a normally distributed population, the sampling distribution of the
sample mean has a normal distribution
a. if n > 30 b. if n/N < .05 c. all the time
8. When samples are selected from a nonnormally distributed population, the sampling distribution of
the sample mean has an approximately normal distribution
a. if n > 30 b. if n/N S .05 c. always
9. In a sample of 200 customers of a mail-order company, 174 are found to be satisfied with the serv-
ice they receive from the company. The proportion of customers in this sample who are satisfied with the
company's service is
a. .87 b. .174 c. .148
10. The mean of the sampling distribution of p is always equal to
a. p b. p c. p
11. The condition for the standard deviation of the sampling distribution of the sample proportion to be
y/pq/n is
a. np > 5 and nq > 5 b. n > 30 c. n/N ' .05
12. The sampling distribution of p is (approximately) normal if
a. np > 5 and nq > 5 b. n > 30 c. n/N < .05
13. Briefly state and explain the central limit theorem.
14. The weights of all students at a large university have an approximately normal distribution with a
mean of 145 pounds and a standard deviation of 18 pounds. Let x be the mean weight of a random sam-
ple of certain students selected from this university. Calculate the mean and standard deviation of x and
describe the shape of its sampling distribution for a sample size of
a. 25 b. 100
15. In a National Highway Transportation Safety Administration (NHTSA) report, data provided to the
NHTSA by Goodyear indicates that the average tread life of properly inflated automobile tires is 45,000
miles (www.nhtsa.dot.gov/cars/rules/rulings/TPMS_FMVSS_Nol38/part5. 5.html). Suppose that the cur-
rent population distribution of tread lives of properly inflated automobile tires has a mean of 45,000 miles
and a standard deviation of 2360 miles, but the shape of this distribution is unknown. Let x be the aver-
age tread life of a random sample of certain properly inflated automobile tires. Calculate the mean and
the standard deviation of the sampling distribution of x and describe its shape for a sample size of
a. 20 b. 65
16. Refer to Problem 15 above. The current population distribution of tread lives of properly inflated au-
tomobile tires has a mean of 45,000 miles and a standard deviation of 2360 miles, but the shape of this
distribution is unknown. Find the probability that the average tread life of a random sample of 65 prop-
erly inflated automobile tires is
a. between 44,500 and 44,750 miles
b. within 180 miles of the population mean
c. 46,000 miles or more
d. not within 400 miles of the population mean
e. less than 44,300 miles
17. At Jen and Perry Ice Cream Company, the machine that fills one-pound cartons of Top Flavor ice
cream is set to dispense 16 ounces of ice cream into every carton. However, some cartons contain
slightly less than and some contain slightly more than 16 ounces of ice cream. The amounts of ice
cream in all such cartons have a normal distribution with a mean of 16 ounces and a standard devia-
tion of .18 ounce.
a. Find the probability that the mean amount of ice cream in a random sample of 16 such cartons
will be
i. between 15.90 and 15.95 ounces
ii. less than 15.95 ounces
iii. more than 15.97 ounces
336
Chapter 7 Sampling Distributions
b. What is the probability that the mean amount of ice cream in a random sample of 16 such cartons
will be within .10 ounce of the population mean?
c. What is the probability that the mean amount of ice cream in a random sample of 16 such cartons
will be less than the population mean by .135 ounce or more?
18. A 2007 article states that 4.8% of U.S. households are "linguistically isolated," which means that all
members of the household aged 14 years and older have difficulty speaking English (Source: http://www.
antara.co.id/en/arc/2007/9/12/five-percent-of-us-families-dont-speak-english-report/). Let p be the proportion
of households in a random sample of U.S. households that are "linguistically isolated." Find the mean and
standard deviation of the sampling distribution of p, and describe its shape when the sample size is
a. 50 b. 500 c. 5000
19. According to the 2008 ARIS American Religious Survey (www.americanreligionsurvey-aris.org),
3.52% of U.S. adults classify themselves as "non-denominational Christians." Assume that this percent-
age is true for the current population of U.S. adults.
a. Find the probability that in a random sample of 900 U.S. adults, the proportion who classify them-
selves as non-denominational Christians is
i. greater than .05
ii. between .03 and .0375
iii. less than .04
iv. between .025 and .0325
b. What is the probability that in a random sample of 900 U.S. adults, the proportion who classify
themselves as non-denominational Christians is within .005 of the population proportion?
c. What is the probability that in a random sample of 900 U.S. adults, the proportion who classify
themselves as non-denominational Christians is not within .008 of the population proportion?
d. What is the probability that in a random sample of 900 U.S. adults, the proportion who classify them-
selves as non-denominational Christians is greater than the population proportion by .0095 or more?
Mini-Projects
■ MINI-PROJECT 7-1
Consider the data on heights of NBA players.
a. Compute p and cr for this data set.
b. Take 20 random samples of five players each, and find x for each sample.
c. Compute the mean and standard deviation of the 20 sample means obtained in part b.
d. Using the formulas given in Section 7.3, find p^ and cr Y for n = 5.
e. How do your values of p^ and Oj in part d compare with those in part c?
f. What percentage of the 20 sample means found in part b lie in the interval p^ — cr- t to p^ + <r{!
In the interval p^ — 2a- K to p- K + 2crj? In the interval p- K — 3cr^ to p- t + 3o"j?
g. How do the percentages in part f compare to the corresponding percentages for a normal distribu-
tion (68%, 95%, and 99.7%, respectively)?
h. Repeat parts b through g using 20 samples of 10 players each.
■ MINI-PROJECT 7-2
Consider Data Set II, Data on States, that accompanies this text. Let p denote the proportion of the 50
states that have a per capita income of less than $35,000.
a. Find p.
b. Select 20 random samples of 5 states each, and find the sample proportion p for each sample.
c. Compute the mean and standard deviation of the 20 sample proportions obtained in part b.
d. Using the formulas given in Section 7.7.2, compute p^ and crp. Is the finite-population correction
factor required here?
e. Compare your mean and standard deviation of p from part c with the values calculated in part d.
f. Repeat parts b through e using 20 samples of 10 states each.
■ MINI-PROJECT 7-3
You are to conduct the experiment of sampling 10 times (with replacement) from the digits 0, 1, 2, 3, 4,
5, 6, 7, 8, and 9. You can do this in a variety of ways. One way is to write each digit on a separate piece
Technology Instruction 337
of paper, place all the slips in a hat, and select 10 times from the hat, returning each selected slip before
the next pick. As alternatives, you can use a 10-sided die, statistical software, or a calculator that generates
random numbers. Perform the experiment using any of these methods, and compute the sample mean x
for the 10 numbers obtained. Now repeat the procedure 49 more times. When you are done, you will have
50 sample means.
a. Make a table of the population distribution for the 10 digits, and display it using a graph.
b. Make a stem-and-leaf display of your 50 sample means. What shape does it have?
c. What does the central limit theorem say about the shape of the sampling distribution of
x? What mean and standard deviation does the sampling distribution of x have in this
problem?
■ MINI-PROJECT 7-4
Reconsider Mini-Project 7-3. Now repeat that project and parts a through c, but this time use a skewed
distribution (as explained below), instead of a symmetric distribution, to take samples. This project is more
easily performed using a computer or graphing calculator, but it can be done using a hat or a random num-
bers table. In this project, sample 10 times from a population of digits that contains twenty 0s, fifteen Is,
ten 2s, seven 3s, four 4s, three 5s, two 6s, and one each of the numbers 7, 8, and 9. Select 50 samples of
size 10, and repeat parts a through c of Mini-Project 7-3. How do the parts a and b of this project com-
pare to parts a and b of Mini-Project 7-3 in regard to the shapes of the distributions? How does this re-
late to what the central limit theorem says?
DECIDE FOR YOURSELF
Deciding About Elections
In the first week of November during an election year, you are very
likely to hear the following statement on TV news, "We are now
able to make a projection. In (insert the name of the state where you
live), we project that the winner will be [insert one of the elected
officials from your state)'' Many people are aware that news agen-
cies conduct exit polls on election day. A commonly asked question
is, "How can an agency call a race based on the results from a sam-
ple of only 1200 voters or so, and do this with a high (although not
perfect) accuracy level?" Although the actual methods used to make
projections based on exit polls are above the level of this book, we will
examine a similar but simpler version of the question here. The con-
cepts and logic involved in this process will help you understand the
statistical inference concepts discussed in subsequent chapters.
Consider a simple election where there are only two candidates,
named A and B. Suppose p and q are the proportions of votes
received by candidates A and B, respectively. Suppose we conduct
an exit poll based on a simple random sample of 800 voters and
determine the mean, standard deviation, and shape of the sampling
distribution of p, where p is the proportion of voters in the sample
who voted for candidate A.
1. Suppose that 440 of the 800 voters included in the exit poll voted
for candidate A, which gives p = .55. Assuming that each candi-
date received 50% of the votes (i.e., p = .50 and q = .50, where p
and q are the proportions of votes received by candidates A and B,
respectively), what is the probability that at least 440 out of 800 vot-
ers in a sample would vote for candidate A?
2. Based on your answer to the above question, the results of the poll
make it reasonable to conclude that the proportion of all voters who
voted for candidate A is actually higher than .5. Explain why.
3. What implications do the above answers have for the result of the
election? Will you make a projection about this election based on the
results of this exit poll?
ECHNOLOGY
INSTRUCTION
Sampling Distribution of Means
1
To create a sampling distribution of a sample mean using the TI-84 requires a good deal of pro-
gramming, which we will not do here. However, it is quite easy to create a sampling distribution
for a sample proportion using the TI-84. Let n and p represent the number of trials and the prob-
ability of a success, respectively, for a binomial experiment. On the TI-84, press 2nd > STAT
>OPS >seq(>MATH >PRB >randBin(n,p)/n,X,l,100) >STO->Ll (see Screen 7.1). This
338 Chapter 7 Sampling Distributions
se*i(randBin(50j .
Xj I j I00)-H_
will produce 100 values of p. If you want more or fewer values of p, change 100 in the above
command to any desired number. Then, you can create a histogram of the data on these p val-
ues using the technology instructions of Chapter 2.
Screen 7.1
Histogram - Simple
□
CI 6
A
C17
CIS
CI 9
C20
C21
C22
C23
C24
C25
C26
C27
C28
C29
C30
C32
V
Graph variables:
Scale...
Labels...
Data View..
Multiple Graphs..
Data Options..
Help
OK
Cancel
1. To see an example of a
sampling distribution of a
sample mean, select Calc
>Random Data >Integer.
We will create 50 samples
of size 30, each value a ran-
dom integer between and
10. Each sample will be a
row, so that when we find
the mean of each row, the
result will go in a column.
2. Enter 50 for Number of
Rows of Data to
Generate.
3. Enter cl-c30 for Store in
columns.
4. Enter for Minimum
value and 10 for Maxi-
mum value.
5. Select OK.
Screen 7.2
Screen 7.3
6. Select Calc >Row Statistics. Select Mean and enter
cl-c30 for Input variables. Enter c32 for Store
results in.
7. Select OK.
8. Select Graph >Histogram. Enter C32 in the Graph
variables: box (see Screen 7.2). Click OK to obtain
the histogram that will appear in the graph window
(see Screen 7.3). Is this histogram bell shaped? What
do you think is the center of this histogram?
9. To see an example of a sampling distribution of a
sample proportion, select Calc >Random Data >
Binomial. Suppose you want to create the number of
successes for 100 binomial experiments, each consist-
ing of 80 trials and a probability of success of .40.
Each row will contain the number of successes for a
set of 80 trials. We will then use these values to cal-
culate the sample proportions for the 100 experiments,
placing the values of the proportions in a different
column. In the dialog box you obtain in response to
the above commands, follow the following steps:
Technology Assignments 339
Enter 100 for Number of Rows of Data to Generate.
Enter CI for Store in columns
Enter 80 for Number of trials and .40 for Event probability.
Click OK.
Select Calc >Calculator. Enter C2 in Store result in variable.
Enter Cl/80 in Expression.
Select OK.
Select Graph >Histogram.
Enter C2 in the Graph variables: box.
Click OK to obtain the histogram that will appear in the graph window. Is this histogram
approximately bell shaped? What do you think is the center of this histogram?
A
B I C
1
3
=average(a1 :a2)
2
8
3
Screen 7.4
1. To see an example of the sampling distribution of means, use the rand function described in
Chapter 1 to create a sample of two random numbers between and 10 in column A.
2. Use the average function described in Chapter 3 to find their mean B. (See Screen 7.4.)
3. Cut and paste the pair of random numbers and their mean 30 times.
4. Use the frequency function described in Chapter 2 to find the frequency counts between
and 1, 1 and 2, 2 and 3, and so forth through 9 and 10.
5. Use the Chart wizard described in Chapter 2 to plot a frequency histogram. Is the his-
togram bell shaped? Where is it centered?
TECHNOLOGY ASSIGNMENTS
TA7.1 Create 200 samples, each containing the results of 30 rolls of a die. Calculate the means of these
200 samples. Construct the histogram, and calculate the mean and standard deviation of these 200 sam-
ple means.
TA7.2 Create 150 samples each containing the results of selecting 35 numbers from 1 through 100. Cal-
culate the means of these 150 samples. Construct the histogram, and calculate the mean and standard de-
viation of these 150 sample means.
TA7.3 Refer to Self-Review Test Exercise 18. In this assignment, we will explore properties of the sam-
pling distribution of a sample proportion for different sample sizes, as well as look at the sample proportion
of failures instead that of successes, Self-Review Test Exercise 18 stated that according to a 2007 article,
4.8% of U.S. households are "linguistically isolated."
a. Using technology, simulate 1000 binomial experiments with 40 trials and a success probability of
.048. Calculate the sample proportion of successes for each of the 1000 experiments. In another
column or list, calculate the sample proportion of failures by subtracting the sample proportion of
successes from 1.
b. Create two histograms (one of the sample proportions of successes, and the other for the sample
proportions of failures). Calculate the mean and standard deviation for each of the sets of 1000
sample proportions. What are the similarities and differences in the histograms and the summary
statistics? Are the histograms approximately bell shaped?
c. Repeat steps a and b with 400 trials. In addition, compare the similarities and differences for 40
and 400 trials.
Chapter
Estimation of the Mean and Proportion
8.1 Estimation: An
Introduction
8.2 Point and Interval
Estimates
8.3 Estimation of a
Population Mean: a
Known
Case Study 8-1
a Child
Raising
8.4 Estimation of a
Population Mean: a Not
Known
8.5 Estimation of a
Population Proportion:
Large Samples
Case Study 8-2 Which Sound
is Most Frustrating
to Hear?
Imagine you are sleeping at night and need to get up early in the morning to go to work. Or,
you are reading something with intense concentration, or watching your most favorite show on TV.
Suddenly a loud noise disturbs you. You get angry. You feel like you have lost something. What is
that frustrating noise that disturbs you? In a survey, 28% of adults said that the most frustrating
sound is the jackhammer. (See Case Study 8-2).
Now we are entering that part of statistics called inferential statistics. In Chapter 1 inferential sta-
tistics was defined as the part of statistics that helps us make decisions about some characteris-
tics of a population based on sample information. In other words, inferential statistics uses the
sample results to make decisions and draw conclusions about the population from which the sam-
ple is drawn. Estimation is the first topic to be considered in our discussion of inferential statistics.
Estimation and hypothesis testing (discussed in Chapter 9) taken together are usually referred to
as inference making. This chapter explains how to estimate the population mean and population
proportion for a single population.
340
8.1 Estimation: An Introduction
341
8.1 Estimation: An Introduction
Estimation is a procedure by which a numerical value or values are assigned to a population
parameter based on the information collected from a sample.
Definition
Estimation The assignment of value(s) to a population parameter based on a value of the
corresponding sample statistic is called estimation.
In inferential statistics, p is called the true population mean and p is called the true
population proportion. There are many other population parameters, such as the median, mode,
variance, and standard deviation.
The following are a few examples of estimation: an auto company may want to estimate
the mean fuel consumption for a particular model of a car; a manager may want to estimate the
average time taken by new employees to learn a job; the U.S. Census Bureau may want to find
the mean housing expenditure per month incurred by households; and the AWAH (Association
of Wives of Alcoholic Husbands) may want to find the proportion (or percentage) of all hus-
bands who are alcoholic.
The examples about estimating the mean fuel consumption, estimating the average time
taken to learn a job by new employees, and estimating the mean housing expenditure per month
incurred by households are illustrations of estimating the true population mean, p. The exam-
ple about estimating the proportion (or percentage) of all husbands who are alcoholic is an il-
lustration of estimating the true population proportion, p.
If we can conduct a census (a survey that includes the entire population) each time we
want to find the value of a population parameter, then the estimation procedures explained
in this and subsequent chapters are not needed. For example, if the U.S. Census Bureau can
contact every household in the United States to find the mean housing expenditure incurred
by households, the result of the survey (which will actually be a census) will give the value
of p, and the procedures learned in this chapter will not be needed. However, it is too ex-
pensive, very time consuming, or virtually impossible to contact every member of a popu-
lation to collect information to find the true value of a population parameter. Therefore, we
usually take a sample from the population and calculate the value of the appropriate sample
statistic. Then we assign a value or values to the corresponding population parameter based
on the value of the sample statistic. This chapter (and subsequent chapters) explains how to
assign values to population parameters based on the values of sample statistics.
For example, to estimate the mean time taken to learn a certain job by new employees,
the manager will take a sample of new employees and record the time taken by each of these
employees to learn the job. Using this information, he or she will calculate the sample mean,
x. Then, based on the value of x, he or she will assign certain values to p. As another ex-
ample, to estimate the mean housing expenditure per month incurred by all households in
the United States, the Census Bureau will take a sample of certain households, collect the
information on the housing expenditure that each of these households incurs per month, and
compute the value of the sample mean, x. Based on this value of x, the bureau will then as-
sign values to the population mean, p. Similarly, the AWAH will take a sample of husbands
and determine the value of the sample proportion, p, which represents the proportion of hus-
bands in the sample who are alcoholic. Using this value of the sample proportion, p, AWAH
will assign values to the population proportion, p.
The value(s) assigned to a population parameter based on the value of a sample statistic is
called an estimate of the population parameter. For example, suppose the manager takes a sam-
ple of 40 new employees and finds that the mean time, x, taken to learn this job for these em-
ployees is 5.5 hours. If he or she assigns this value to the population mean, then 5.5 hours is
called an estimate of p. The sample statistic used to estimate a population parameter is called
an estimator. Thus, the sample mean, x, is an estimator of the population mean, p; and the
sample proportion, p, is an estimator of the population proportion, p.
342
Chapter 8 Estimation of the Mean and Proportion
Definition
Estimate and Estimator The value(s) assigned to a population parameter based on the value of
a sample statistic is called an estimate. The sample statistic used to estimate a population param-
eter is called an estimator.
The estimation procedure involves the following steps.
1. Select a sample.
2. Collect the required information from the members of the sample.
3. Calculate the value of the sample statistic.
4. Assign value(s) to the corresponding population parameter.
Remember, the procedures to be learned in this chapter assume that the sample taken
is a simple random sample. If the sample is not a simple random sample (see Appendix A for
a few other kinds of samples), then the procedures to be used to estimate a population mean or
proportion become more complex. These procedures are outside the scope of this book.
8.2 Point and Interval Estimates
An estimate may be a point estimate or an interval estimate. These two types of estimates are
described in this section.
8.2.1 A Point Estimate
If we select a sample and compute the value of the sample statistic for this sample, then this
value gives the point estimate of the corresponding population parameter.
Definition
Point Estimate The value of a sample statistic that is used to estimate a population parameter is
called a point estimate.
Thus, the value computed for the sample mean, x, from a sample is a point estimate of the
corresponding population mean, fi. For the example mentioned earlier, suppose the Census
Bureau takes a sample of 10,000 households and determines that the mean housing expenditure
per month, x, for this sample is $1970. Then, using x as a point estimate of fi, the Bureau can
state that the mean housing expenditure per month, /jl, for all households is about $1970. Thus,
Point estimate of a population parameter = Value of the corresponding sample statistic
Each sample selected from a population is expected to yield a different value of the sam-
ple statistic. Thus, the value assigned to a population mean, fi, based on a point estimate de-
pends on which of the samples is drawn. Consequently, the point estimate assigns a value to fi
that almost always differs from the true value of the population mean.
8.2.2 An Interval Estimate
In the case of interval estimation, instead of assigning a single value to a population parameter,
an interval is constructed around the point estimate, and then a probabilistic statement that this
interval contains the corresponding population parameter is made.
Definition
Interval Estimation In interval estimation, an interval is constructed around the point estimate,
and it is stated that this interval is likely to contain the corresponding population parameter.
8.2 Point and Interval Estimates 343
For the example about the mean housing expenditure, instead of saying that the mean
housing expenditure per month for all households is $1970, we may obtain an interval by sub-
tracting a number from $1970 and adding the same number to $1970. Then we state that this
interval contains the population mean, p. For purposes of illustration, suppose we subtract
$340 from $1970 and add $340 to $1970. Consequently, we obtain the interval ($1970 - $340)
to ($1970 + $340), or $1630 to $2310. Then we state that the interval $1630 to $2310 is
likely to contain the population mean, p, and that the mean housing expenditure per month
for all households in the United States is between $1630 and $2310. This procedure is called
interval estimation. The value $1630 is called the lower limit of the interval, and $2310 is
called the upper limit of the interval. The number we add to and subtract from the point es-
timate is called the margin of error. Figure 8.1 illustrates the concept of interval estimation.
Figure 8.1 Interval estimation.
$1630 $2310
The question arises: What number should we subtract from and add to a point estimate to
obtain an interval estimate? The answer to this question depends on two considerations:
1. The standard deviation <x- of the sample mean, x
2. The level of confidence to be attached to the interval
First, the larger the standard deviation of I, the greater is the number subtracted from and
added to the point estimate. Thus, it is obvious that if the range over which x can assume val-
ues is larger, then the interval constructed around x must be wider to include p.
Second, the quantity subtracted and added must be larger if we want to have a higher con-
fidence in our interval. We always attach a probabilistic statement to the interval estimation.
This probabilistic statement is given by the confidence level. An interval constructed based on
this confidence level is called a confidence interval.
Definition
Confidence Level and Confidence Interval Each interval is constructed with regard to a given
confidence level and is called a confidence interval. The confidence interval is given as
Point estimate ± Margin of error
The confidence level associated with a confidence interval states how much confidence we have
that this interval contains the true population parameter. The confidence level is denoted by
(1 - a)100%.
The confidence level is denoted by (1 — a)100%, where a is the Greek letter alpha. When
expressed as probability, it is called the confidence coefficient and is denoted by 1 — a. In pass-
ing, note that a is called the significance level, which will be explained in detail in Chapter 9.
Although any value of the confidence level can be chosen to construct a confidence interval,
the more common values are 90%, 95%, and 99%. The corresponding confidence coefficients
are .90, .95, and .99, respectively. The next section describes how to construct a confidence
interval for the population mean when the population standard deviation, cr, is known.
Sections 8.3 and 8.4 discuss the procedures that are used to estimate a population mean p.
In Section 8.3 we assume that the population standard deviation cr is known, and in Section 8.4
we do not assume that the population standard deviation cr is known. In the latter situation, we
344
Chapter 8 Estimation of the Mean and Proportion
use the sample standard deviation s instead of cr. In the real world, the population standard de-
viation cr is almost never known. Consequently, we (almost) always use the sample standard
deviation s.
8.3 Estimation of a Population Mean: a Known
This section explains how to construct a confidence interval for the population mean j± when
the population standard deviation cr is known. Here, there are three possible cases, as follows.
Case I. If the following three conditions are fulfilled:
1. The population standard deviation cr is known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is normally distributed,
then we use the normal distribution to make the confidence interval for /jl because from Section 7.4. 1
of Chapter 7 the sampling distribution of x is normal with its mean equal to /jl and the standard
deviation equal to cr^ = cr/vn, assuming that n/N s .05.
Case II. If the following two conditions are fulfilled:
1. The population standard deviation cr is known
2. The sample size is large (i.e., n > 30),
then, again, we use the normal distribution to make the confidence interval for /jl because from
Section 7.4.2 of Chapter 7, due to the central limit theorem, the sampling distribution of x is
(approximately) normal with its mean equal to /jl and the standard deviation equal to cr T = cr/ Vra,
assuming that n/N s .05.
Case III. If the following three conditions are fulfilled:
1. The population standard deviation cr is known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is not normally distributed (or its
distribution is unknown),
then we use a nonparametric method to make the confidence interval for /jl. Such procedures
are covered in Chapter 15 that is on the Web site of the text.
This section will cover the first two cases. The procedure for making a confidence interval
for j± is the same in both these cases. Note that in Case I, the population does not have to be
exactly normally distributed. As long as it is close to the normal distribution without any out-
liers, we can use the normal distribution procedure. In Case II, although 30 is considered a large
sample, if the population distribution is very different from the normal distribution, then 30 may
not be a large enough sample size for the sampling distribution of x to be normal and, hence,
to use the normal distribution.
The following chart summarizes the above three cases.
cr Is Known
1
I
Case I
1. n < 30
2. Population is normal
Case II
n 2: 30
Casein
1. n < 30
2. Population is not normal
Use the normal distribution
to estimate /a
Use a nonparametric method
to estimate
8.3 Estimation of a Population Mean: <r Known 345
Confidence Interval for ft The (1 — a)100% confidence interval for /jl under Cases I and II
is
I ± zo-j
where cr T = cr/Vn
The value of z used here is obtained from the standard normal distribution table (Table IV of
Appendix C) for the given confidence level.
The quantity zaj in the confidence interval formula is called the margin of error and is
denoted by E.
Definition
Margin of Error The margin of error for the estimate for /j,, denoted by E, is the quantity that is
subtracted from and added to the value of x to obtain a confidence interval for /jl. Thus,
E = za x
The value of z in the confidence interval formula is obtained from the standard normal dis-
tribution table (Table IV of Appendix C) for the given confidence level. To illustrate, suppose
we want to construct a 95% confidence interval for fi. A 95% confidence level means that the
total area under the normal curve for x between two points (at the same distance) on different
sides of jj, is 95%, or .95, as shown in Figure 8.2. Note that we have denoted these two points
by Z\ and z 2 m Figure 8.2. To find the value of z for a 95% confidence level, we first find the
areas to the left of these two points, Zi and z 2 . Then we find the z values for these two areas
from the normal distribution table. Note that these two values of z will be the same but with
opposite signs. To find these values of z, we perform the following two steps:
Figure 8.2 Finding z for a 95% confidence level.
1. The first step is to find the areas to the left of Zi and z 2 , respectively. Note that the area be-
tween Z\ and z 2 is denoted by 1 — a. Hence, the total area in the two tails is a because the
total area under the curve is 1.0. Therefore, the area in each tail, as shown in Figure 8.3, is
a/2. In our example, 1 — a = .95. Hence, the total area in both tails is a = 1 — .95 = .05.
Consequently, the area in each tail is a/2 = .05/2 = .025. Then, the area to the left of Z\ is
.0250, and the area to the left of z 2 is .0250 + .95 = .9750.
2. Now find the z values from Table IV of Appendix C such that the areas to the left of Z\ and
z 2 are .0250 and .9750, respectively. These z values are —1.96 and 1.96, respectively.
Thus, for a confidence level of 95%, we will use z = 1.96 in the confidence interval formula.
a
/ (1 -a)
a
2
I 2
■
I
Figure 8.3 Area in the tails.
-z z z
346 Chapter 8 Estimation of the Mean and Proportion
Table 8.1 lists the z values for some of the most commonly used confidence levels. Note
that we always use the positive value of z in the formula.
Table 8.1 z Values for Commonly Used Confidence Levels
Confidence Level
Areas to Look for in Table IV
z Value
90%
.0500 and .9500
1.64 or 1.65
95%
.0250 and .9750
1.96
96%
.0200 and .9800
2.05
97%
.0150 and .9850
2.17
98%
.0100 and .9900
2.33
99%
.0050 and .9950
2.57 or 2.58
Example 8-1 describes the procedure used to construct a confidence interval for j± when
cr is known, the sample size is small, but the population from which the sample is drawn is nor-
mally distributed.
EXAMPLE 8-1
Finding the point estimate
and confidence interval for
ft: a known, n < 30, and
population normal.
A publishing company has just published a new college textbook. Before the company decides
the price at which to sell this textbook, it wants to know the average price of all such text-
books in the market. The research department at the company took a sample of 25 compara-
ble textbooks and collected information on their prices. This information produced a mean
price of $145 for this sample. It is known that the standard deviation of the prices of all such
textbooks is $35 and the population of such prices is normal.
(a) What is the point estimate of the mean price of all such college textbooks?
(b) Construct a 90% confidence interval for the mean price of all such college textbooks.
Solution Here, a is known and, although n < 30, the population is normally distributed.
Hence, we can use the normal distribution. From the given information,
n = 25,
The standard deviation of x is
$145, and a = $35
cr
Vn
35
25
$7.00
(a) The point estimate of the mean price of all such college textbooks is $145; that is,
Point estimate of /x = x = $145
(b) The confidence level is 90%, or .90. First we find the z value for a 90% confidence level.
Here, the area in each tail of the normal distribution curve is a/2 = (1 — .90)/2 = .05.
Now in Table IV, look for the areas .0500 and .9500 and find the corresponding val-
ues of z. These values are z = —1.65 and z = 1.65.
Next, we substitute all the values in the confidence interval formula for fi. The
90% confidence interval for /jl is
jc ± zo- 7x = 145 ± 1.65(7.00) = 145 ± 11.55
= (145 - 11.55) to (145 + 11.55) = $133.45 to $156.55
Thus, we are 90% confident that the mean price of all such college textbooks is
between $133.45 and $156.55. Note that we cannot say for sure whether the interval
'Note that there is no apparent reason for choosing .0505 and .9505, and not choosing .0495 and .9495. If we choose
.0495 and .9495, the z values will be —1.64 and 1.64. An alternative is to use the average of 1.64 and 1.65, which is
1.645, which we will not do in this text.
8.3 Estimation of a Population Mean: a Known 347
$133.45 to $156.55 contains the true population mean or not. Because fi is a constant,
we cannot say that the probability is .90 that this interval contains because either it
contains /x or it does not. Consequently, the probability is either 1.0 or that this in-
terval contains /jl. All we can say is that we are 90% confident that the mean price of
all such college textbooks is between $133.45 and $156.55.
In the above estimate, $11.55 is called the margin of error or give-and-take
figure. I
How do we interpret a 90% confidence level? In terms of Example 8-1, if we take all possi-
ble samples of 25 such college textbooks each and construct a 90% confidence interval for /jl
around each sample mean, we can expect that 90% of these intervals will include /jl and 10% will
not. In Figure 8.4 we show means x x , x 2 , and x 3 of three different samples of the same size drawn
from the same population. Also shown in this figure are the 90% confidence intervals constructed
around these three sample means. As we observe, the 90% confidence intervals constructed around
Xi and x 2 include jx, but the one constructed around x 3 does not. We can state for a 90% confi-
dence level that if we take many samples of the same size from a population and construct 90%
confidence intervals around the means of these samples, then 90% of these confidence intervals
will be like the ones around 3tj and x 2 in Figure 8.4, which include /x, and 10% will be like the
one around x 3 , which does not include jju.
Example 8-2 illustrates how to obtain a confidence interval for /j, when cr is known and
the sample size is large (« > 30).
■ EXAMPLE 8-2
In a 2009 survey by I-Pension LLC, adults with annual household incomes of $50,000 to
$125,000 were asked about the average time they spend reviewing their 401(k) statements. Of
Constructing a confidence
interval for fi: a known and
the adults surveyed, about 72% said that they spend less than 5 minutes, and 27% said that > 3fl
they spend 5 to 10 minutes to review their 401(k) statements (USA TODAY, February 16, 2009).
Suppose a random sample of 400 adults of all income levels who have 401(k) accounts were
recently asked about the times they spend reviewing their 401(k) statements. The sample pro-
duced a mean of 8 minutes. Assume that the standard deviation of such times for all 401(k)
account holders is 2.20 minutes. Construct a 99% confidence interval for the mean time spent
by all 401(k) account holders reviewing their 401(k) statements.
Solution From the given information,
n = 400, x = 8 minutes, cr = 2.20 minutes,
and Confidence level = 99%, or .99
In this example, we know the population standard deviation cr. Although the shape of the pop-
ulation distribution is unknown, the population standard deviation is known, and the sample
348 Chapter 8 Estimation of the Mean and Proportion
size is large (n > 30). Hence, we can use the normal distribution to make a confidence inter-
val for j±. To make this confidence interval, first we find the standard deviation of x. The value
of o-y is
rr 1 70
-- .11
To find z for a 99% confidence level, first we find area in each of the two tails of the normal
distribution curve, which is (1 - -99)/2 = .0050. Then, we look for .0050 and .0050 +
.99 = .9950 areas in the normal distribution table to find the two z values. These two z values
are (approximately) —2.58 and 2.58. Thus, we will use z = 2.58 in the confidence interval for-
mula. Substituting all the values in the formula, we obtain the 99% confidence interval for /j,,
x ± za- 7x = 8 ± 2.58(.ll) = 8 ± .28 = 7.72 to 8.28 minutes
Thus, we can state with 99% confidence that the mean time spent by all 401(k) account
holders reviewing their 401(k) statements is between 7.72 and 8.28 minutes. I
The width of a confidence interval depends on the size of the margin of error, zoj, which
depends on the values of z, cr, and n because <Tj. = cr/ Vn. However, the value of cr is not under
the control of the investigator. Hence, the width of a confidence interval can be controlled using
1. The value of z, which depends on the confidence level
2. The sample size n
The confidence level determines the value of z, which in turn determines the size of the
margin of error. The value of z increases as the confidence level increases, and it decreases as
the confidence level decreases. For example, the value of z is approximately 1.65 for a 90%
confidence level, 1.96 for a 95% confidence level, and approximately 2.58 for a 99% confidence
level. Hence, the higher the confidence level, the larger the width of the confidence interval,
other things remaining the same.
For the same value of cr, an increase in the sample size decreases the value of cr-, which in
turn decreases the size of the margin of error when the confidence level remains unchanged.
Therefore, an increase in the sample size decreases the width of the confidence interval.
Thus, if we want to decrease the width of a confidence interval, we have two choices:
1. Lower the confidence level
2. Increase the sample size
Lowering the confidence level is not a good choice, however, because a lower confidence level
may give less reliable results. Therefore, we should always prefer to increase the sample size if
we want to decrease the width of a confidence interval. Next we illustrate, using Example 8-2,
how either a decrease in the confidence level or an increase in the sample size decreases the
width of the confidence interval.
O Confidence Level and the Width of the Confidence Interval
Reconsider Example 8-2. Suppose all the information given in that example remains the same.
First, let us decrease the confidence level to 95%. From the normal distribution table, z = 1.96
for a 95% confidence level. Then, using z = 1.96 in the confidence interval for Example 8-2,
we obtain
x ± za-j. = 8 ± 1.96(.ll) = 8 ± .22 = 7.78 to 8.22
Comparing this confidence interval to the one obtained in Example 8-2, we observe that
the width of the confidence interval for a 95% confidence level is smaller than the one for a
99% confidence level.
Sample Size and the Width of the Confidence Interval
Consider Example 8-2 again. Now suppose the information given in that example is based on
a sample size of 1600. Further assume that all other information given in that example, including
USA TODAY Snapshots®
Raising a child
What families will spend to raise a
child bom in 2008 through age 17 1 :
Family income/Cost:
1 - 2008 dollars
Source: USDA*s center for Nutrition Policy and Promotion
RAISING A
CHILD
By Anne R, Carey and Alejandro Gonzalez, USA TODAY
The above chart, based on the U.S. Department of Agriculture's Center for Nutrition Policy and Promotion
study, gives the average cost of raising a child born in 2008 for three different income groups. For exam-
ple, families with annual incomes less than $56,870 are expected to spend an average of $159,870 on their
child born in 2008 through age 17. The corresponding average expenditures for families with annual in-
comes of $56,870 to $98,470 and families with annual incomes of more than $98,470 will be $221,190
and $366,660, respectively. Of course, these averages are based on sample surveys of families. If we know
the sample sizes and the population standard deviations for these three groups, we can then find the con-
fidence interval for the mean expenditure incurred by each group on a child born in 2008 through age 17
as shown in the following table.
Family Income Mean Expenditure Confidence Interval
Less than $56,870 $159,870 $159,870 ± zo-j
$56,870 to $98,470 $221,190 $221,190 ± za- x
More than $98,470 $366,660 $366,660 ± zo-j
For each confidence interval listed in the table, we can substitute the value of z and the value of cr a which
is calculated as For example, suppose we want to find a 98% confidence interval for the mean
Vn
expenditure on a child by families with annual incomes of $56,870 to $98,470. Assume that the mean
expenditure of $221,190 for this income group given in the chart is based on a random sample of 3600
such families and that the population standard deviation for such expenditures for families in this income
group is $30,000. Then the 98% confidence interval for the corresponding population mean is calculated
as follows.
Vn V3600
x ± zo-j = 221,190 ± 2.33(500) = 221,190 ± 1 1 65 = $220,025 to $222,355
We can find the confidence intervals for the population means of the other two groups mentioned in
the chart and table the same way. Note that the sample means given in the table are the point estimates
Source: The chart is reproduced with
permission from USA TODAY, August 21
of the corresponding population means. 2009. Copyright © 2009, USA TODAY.
349
350 Chapter 8 Estimation of the Mean and Proportion
the confidence level, remains the same. First, we calculate the standard deviation of the sample
mean using n = 1600:
ctj = a/Vn = 2.20/VT600 = .055
Then, the 99% confidence interval for fx. is
x ± z<t 7x = 8 ± 2.58(.055) = 8 ± .14 = 7.86 to 8.14
Comparing this confidence interval to the one obtained in Example 8-2, we observe that
the width of the 99% confidence interval for n = 1600 is smaller than the 99% confidence
interval for n = 400.
8.3.1 Determining the Sample Size for the Estimation of Mean
One reason we usually conduct a sample survey and not a census is that almost always we have
limited resources at our disposal. In light of this, if a smaller sample can serve our purpose,
then we will be wasting our resources by taking a larger sample. For instance, suppose we want
to estimate the mean life of a certain auto battery. If a sample of 40 batteries can give us the
confidence interval we are looking for, then we will be wasting money and time if we take a
sample of a much larger size — say, 500 batteries. In such cases, if we know the confidence level
and the width of the confidence interval that we want, then we can find the (approximate) size
of the sample that will produce the required result.
From earlier discussion, we learned that E = z(T- K is called the margin of error of estimate
for fx. As we know, the standard deviation of the sample mean is equal to cr/vn. Therefore,
we can write the margin of error of estimate for fi as
<7
Suppose we predetermine the size of the margin of error, E, and want to find the size of
the sample that will yield this margin of error. From the above expression, the following for-
mula is obtained that determines the required sample size n.
Determining the Sample Size for the Estimation of n Given the confidence level and the stan-
dard deviation of the population, the sample size that will produce a predetermined margin of
error E of the confidence interval estimate of /jl is
If we do not know cr, we can take a preliminary sample (of any arbitrarily determined size)
and find the sample standard deviation, s. Then we can use s for cr in the formula. However,
note that using s for cr may give a sample size that eventually may produce an error much larger
(or smaller) than the predetermined margin of error. This will depend on how close s and cr are.
Example 8-3 illustrates how we determine the sample size that will produce the margin of
error of estimate for /jl wihin a certain limit.
Determining the sample size for
the estimation of fx,.
■ EXAMPLE 8-3
An alumni association wants to estimate the mean debt of this year's college graduates. It is
known that the population standard deviation of the debts of this year's college graduates is
$11,800. How large a sample should be selected so that the estimate with a 99% confidence
level is within $800 of the population mean?
Solution The alumni association wants the 99% confidence interval for the mean debt of
this year's college graduates to be
x ± 800
8.3 Estimation of a Population Mean: a Known 351
Hence, the maximum size of the margin of error of estimate is to be $800; that is,
E = $800
The value of z for a 99% confidence level is 2.58. The value of a is given to be $11,800.
Therefore, substituting all values in the formula and simplifying, we obtain
(2.58) 2 (11,800) 2
2 2
1448.18 » 1449
E 2 (800) 2
Thus, the required sample size is 1449. If the alumni association takes a sample of 1449
of this year's college graduates, computes the mean debt for this sample, and then makes a
99% confidence interval around this sample mean, the margin of error of estimate will be
approximately $800. Note that we have rounded the final answer for the sample size to the
next higher integer. This is always the case when determining the sample size. H
e. 98% f. 99%
= 24.5. It is known that a
EXERCISES
CONCEPTS AND PROCEDURES
8.1 Briefly explain the meaning of an estimator and an estimate.
8.2 Explain the meaning of a point estimate and an interval estimate.
8.3 What is the point estimator of the population mean, xi? How would you calculate the margin of er-
ror for an estimate of /x?
8.4 Explain the various alternatives for decreasing the width of a confidence interval. Which is the best
alternative?
8.5 Briefly explain how the width of a confidence interval decreases with an increase in the sample size.
Give an example.
8.6 Briefly explain how the width of a confidence interval decreases with a decrease in the confidence
level. Give an example.
8.7 Briefly explain the difference between a confidence level and a confidence interval.
8.8 What is the margin of error of estimate for /x when cr is known? How is it calculated?
8.9 How will you interpret a 99% confidence interval for /x? Explain.
8.10 Find z for each of the following confidence levels,
a. 90% b. 95% c. 96% d. 97%
8.11 For a data set obtained from a sample, n = 20 and x
lation is normally distributed.
a. What is the point estimate of /x?
b. Make a 99% confidence interval for /x.
c. What is the margin of error of estimate for part b?
8.12 For a data set obtained from a sample, n = 81 and x = 48.25. It is known that cr = 4.8.
a. What is the point estimate of /x?
b. Make a 95% confidence interval for /x.
c. What is the margin of error of estimate for part b?
8.13 The standard deviation for a population is cr = 15.3. A sample of 36 observations selected from this
population gave a mean equal to 74.8.
a. Make a 90% confidence interval for /jl.
b. Construct a 95% confidence interval for \l.
c. Determine a 99% confidence interval for /x.
d. Does the width of the confidence intervals constructed in parts a through c increase as the con-
fidence level increases? Explain your answer.
8.14 The standard deviation for a population is cr = 14.8. A sample of 25 observations selected from this
population gave a mean equal to 143.72. The population is known to have a normal distribution.
a. Make a 99% confidence interval for /jl.
b. Construct a 95% confidence interval for \l.
c. Determine a 90% confidence interval for /x.
d. Does the width of the confidence intervals constructed in parts a through c decrease as the con-
fidence level decreases? Explain your answer.
3.1. The popu-
352 Chapter 8 Estimation of the Mean and Proportion
8.15 The standard deviation for a population is cr = 6.30. A random sample selected from this popula-
tion gave a mean equal to 81.90. The population is known to be normally distributed.
a. Make a 99% confidence interval for /x assuming n = 16.
b. Construct a 99% confidence interval for /x assuming n = 20.
c. Determine a 99% confidence interval for /x assuming n = 25.
d. Does the width of the confidence intervals constructed in parts a through c decrease as the sample
size increases? Explain.
8.16 The standard deviation for a population is cr = 7.14. A random sample selected from this popula-
tion gave a mean equal to 48.52.
a. Make a 95% confidence interval for /x assuming n = 196.
b. Construct a 95% confidence interval for /x assuming n = 100.
c. Determine a 95% confidence interval for /x assuming n = 49.
d. Does the width of the confidence intervals constructed in parts a through c increase as the sam-
ple size decreases? Explain.
8.17 For a population, the value of the standard deviation is 2.65. A sample of 35 observations taken from
this population produced the following data.
42
51
42
31
28
36
49
29
46
37
32
27
33
41
47
41
28
46
34
39
48
26
35
37
38
46
48
39
29
31
44
41
37
38
46
a. What is the point estimate of /x?
b. Make a 98% confidence interval for (i.
c. What is the margin of error of estimate for part b?
8.18 For a population, the value of the standard deviation is 4.96. A sample of 32 observations taken from
this population produced the following data.
74
85
72
73
86
81
77
60
83
78
79
88
76
73
84
78
81
72
82
81
79
83
88
86
78
83
87
82
80
84
76
74
a. What is the point estimate of /x?
b. Make a 99% confidence interval for /x.
c. What is the margin of error of estimate for part b?
8.19 For a population data set, cr = 12.5.
a. How large a sample should be selected so that the margin of error of estimate for a 99% confi-
dence interval for xx is 2.50?
b. How large a sample should be selected so that the margin of error of estimate for a 96% confi-
dence interval for /x is 3.20?
8.20 For a population data set, cr = 14.50.
a. What should the sample size be for a 98% confidence interval for /x to have a margin of error of
estimate equal to 5.50?
b. What should the sample size be for a 95% confidence interval for /x to have a margin of error of
estimate equal to 4.25?
8.21 Determine the sample size for the estimate of /x for the following.
a. E = 2.3, cr = 15.40, confidence level = 99%
b. E = 4.1, cr = 23.45, confidence level = 95%
c. E = 25.9, cr = 122.25, confidence level = 90%
8.22 Determine the sample size for the estimate of /x for the following.
a. E = .17, cr = .90, confidence level = 99%
b. E = 1.45, cr = 5.82, confidence level = 95%
c. E = 5.65, cr = 18.20, confidence level = 90%
■ APPLICATIONS
8.23 A sample of 1500 homes sold recently in a state gave the mean price of homes equal to $299,720.
The population standard deviation of the prices of homes in this state is $68,650. Construct a 99% confi-
dence interval for the mean price of all homes in this state.
8.3 Estimation of a Population Mean: a Known
353
8.24 Inside the Box Corporation makes corrugated cardboard boxes. One type of these boxes states that
the breaking capacity of this box is 75 pounds. Fifty-five randomly selected such boxes were loaded un-
til they broke. The average breaking capacity of these boxes was found to be 78.52 pounds. Suppose that
the standard deviation of the breaking capacities of all such boxes is 2.63 pounds. Calculate a 99% con-
fidence interval for the average breaking capacity of all boxes of this type.
8.25 KidPix Entertainment is in the planning stages of producing a new computer-animated movie for na-
tional release, so they need to determine the production time (labor-hours necessary) to produce the movie.
The mean production time for a random sample of 14 big-screen computer-animated movies is found to
be 53,550 labor-hours. Suppose that the population standard deviation is known to be 7462 labor-hours
and the distribution of production times is normal.
a. Construct a 98% confidence interval for the mean production time to produce a big-screen
computer-animated movie.
b. Explain why we need to make the confidence interval. Why is it not correct to say that the average
production time needed to produce all big-screen computer-animated movies is 53,550 labor-hours?
8.26 Lazurus Steel Corporation produces iron rods that are supposed to be 36 inches long. The machine
that makes these rods does not produce each rod exactly 36 inches long. The lengths of the rods vary
slightly. It is known that when the machine is working properly, the mean length of the rods made on this
machine is 36 inches. The standard deviation of the lengths of all rods produced on this machine is al-
ways equal to .10 inch. The quality control department takes a sample of 20 such rods every week, cal-
culates the mean length of these rods, and makes a 99% confidence interval for the population mean. If
either the upper limit of this confidence interval is greater than 36.05 inches or the lower limit of this con-
fidence interval is less than 35.95 inches, the machine is stopped and adjusted. A recent sample of 20 rods
produced a mean length of 36.02 inches. Based on this sample, will you conclude that the machine needs
an adjustment? Assume that the lengths of all such rods have a normal distribution.
8.27 At Farmer's Dairy, a machine is set to fill 32-ounce milk cartons. However, this machine does not put
exactly 32 ounces of milk into each carton; the amount varies slightly from carton to carton. It is known that
when the machine is working properly, the mean net weight of these cartons is 32 ounces. The standard de-
viation of the amounts of milk in all such cartons is always equal to .15 ounce. The quality control depart-
ment takes a sample of 25 such cartons every week, calculates the mean net weight of these cartons, and makes
a 99% confidence interval for the population mean. If either the upper limit of this confidence interval is greater
than 32.15 ounces or the lower limit of this confidence interval is less than 31.85 ounces, the machine is
stopped and adjusted. A recent sample of 25 such cartons produced a mean net weight of 31.94 ounces. Based
on this sample, will you conclude that the machine needs an adjustment? Assume that the amounts of milk
put in all such cartons have a normal distribution.
8.28 A consumer agency that proposes that lawyers' rates are too high wanted to estimate the mean hourly
rate for all lawyers in New York City. A sample of 70 lawyers taken from New York City showed that the
mean hourly rate charged by them is $420. The population standard deviation of hourly charges for all
lawyers in New York City is $110.
a. Construct a 99% confidence interval for the mean hourly charges for all lawyers in New York City.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this inter-
val be reduced? Discuss all possible alternatives. Which alternative is the best?
8.29 A bank manager wants to know the mean amount of mortgage paid per month by homeowners in
an area. A random sample of 120 homeowners selected from this area showed that they pay an average
of $1575 per month for their mortgages. The population standard deviation of such mortgages is $215.
a. Find a 97% confidence interval for the mean amount of mortgage paid per month by all home-
owners in this area.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this inter-
val be reduced? Discuss all possible alternatives. Which alternative is the best?
8.30 A marketing researcher wants to find a 95% confidence interval for the mean amount that visitors
to a theme park spend per person per day. She knows that the standard deviation of the amounts spent per
person per day by all visitors to this park is $11. How large a sample should the researcher select so that
the estimate will be within $2 of the population mean?
8.31 A company that produces detergents wants to estimate the mean amount of detergent in 64-ounce
jugs at a 99% confidence level. The company knows that the standard deviation of the amounts of deter-
gent in all such jugs is .20 ounce. How large a sample should the company select so that the estimate is
within .04 ounce of the population mean?
8.32 A department store manager wants to estimate at a 90% confidence level the mean amount spent by all
customers at this store. The manager knows that the standard deviation of amounts spent by all customers at
this store is $31. What sample size should he choose so that the estimate is within $3 of the population mean?
354
Chapter 8 Estimation of the Mean and Proportion
8.33 The principal of a large high school is concerned about the amount of time that his students spend
on jobs to pay for their cars, to buy clothes, and so on. He would like to estimate the mean number of
hours worked per week by these students. He knows that the standard deviation of the times spent per week
on such jobs by all students is 2.5 hours. What sample size should he choose so that the estimate is within
.75 hour of the population mean? The principal wants to use a 98% confidence level.
*8.34 You are interested in estimating the mean commuting time from home to school for all commuter
students at your school. Briefly explain the procedure you will follow to conduct this study. Collect the
required data from a sample of 30 or more such students and then estimate the population mean at a 99%
confidence level. Assume that the population standard deviation for such times is 5.5 minutes.
*8.35 You are interested in estimating the mean age of cars owned by all people in the United States.
Briefly explain the procedure you will follow to conduct this study. Collect the required data on a sample
of 30 or more cars and then estimate the population mean at a 95% confidence level. Assume that the pop-
ulation standard deviation is 2.4 years.
8.4 Estimation of a Population Mean:
a Not Known
This section explains how to construct a confidence interval for the population mean fi when
the population standard deviation a is not known. Here, again, there are three possible cases
that are mentioned below.
Case I. If the following three conditions are fulfilled:
1. The population standard deviation a is not known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is normally distributed,
then we use the t distribution (explained in Section 8.4.1) to make the confidence interval for /jl.
Case II. If the following two conditions are fulfilled:
1. The population standard deviation a is not known
2. The sample size is large (i.e., n > 30),
then again we use the t distribution to make the confidence interval for jx.
Case III. If the following three conditions are fulfilled:
1. The population standard deviation a is not known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is not normally distributed (or its
distribution is unknown),
then we use a nonparametric method to make the confidence interval for /j,. Such procedures
are covered in Chapter 15, which is on the Web site of this text.
The following chart summarizes the above three cases.
CT Is Not Known
1
I
1
Case I
1. n < 30
2. Population is normal
Case II
n > 30
Case III
1. n < 30
2. Population is not normal
Use the r distribution to
estimate /a
Use a nonparametric method
to estimate /j.
8.4 Estimation of a Population Mean: <x Not Known 355
In the next subsection, we discuss the t distribution, and then in Section 8.4.2 we show how
to use the t distribution to make a confidence interval for /jl when cr is not known and condi-
tions of Cases I and II are satisfied.
8.4.1 The t Distribution
The t distribution was developed by W. S. Gosset in 1908 and published under the pseudo-
nym Student. As a result, the t distribution is also called Student's t distribution. The t distri-
bution is similar to the normal distribution in some respects. Like the normal distribution curve,
the t distribution curve is symmetric (bell shaped) about the mean and never meets the hori-
zontal axis. The total area under a t distribution curve is 1 .0, or 100%. However, the t distribution
curve is flatter than the standard normal distribution curve. In other words, the t distribu-
tion curve has a lower height and a wider spread (or, we can say, a larger standard deviation)
than the standard normal distribution. However, as the sample size increases, the t distribution
approaches the standard normal distribution. The units of a t distribution are denoted by t.
The shape of a particular t distribution curve depends on the number of degrees of free-
dom (df). For the purpose of Chapters 8 and 9, the number of degrees of freedom for a t dis-
tribution is equal to the sample size minus one, that is,
df = n — 1
The number of degrees of freedom is the only parameter of the t distribution. There is a different
t distribution for each number of degrees of freedom. Like the standard normal distribution, the
mean of the t distribution is 0. But unlike the standard normal distribution, whose standard de-
viation is 1, the standard deviation of a t distribution is \ / df/(df — 2), which is always greater
than 1. Thus, the standard deviation of a t distribution is larger than the standard deviation of
the standard normal distribution.
Definition
The t Distribution The t distribution is a specific type of bell-shaped distribution with a lower
height and a wider spread than the standard normal distribution. As the sample size becomes
larger, the t distribution approaches the standard normal distribution. The t distribution has only
one parameter, called the degrees of freedom (df). The mean of the t distribution is equal to 0,
and its standard deviation is \'df/(df — 2).
Figure 8.5 shows the standard normal distribution and the t distribution for 9 degrees of
freedom. The standard deviation of the standard normal distribution is 1 .0, and the standard de-
viation of the t distribution is Vdf/(df - 2) = V9/(9 - 2) = 1.134.
Figure 8.5 The t distribution for df = 9 and
the standard normal distribution.
fi =
As stated earlier, the number of degrees of freedom for a t distribution for the purpose of
this chapter is n — 1. The number of degrees of freedom is defined as the number of observa-
tions that can be chosen freely. As an example, suppose we know that the mean of four values
is 20. Consequently, the sum of these four values is 20(4) = 80. Now, how many values out of
356 Chapter 8 Estimation of the Mean and Proportion
four can we choose freely so that the sum of these four values is 80? The answer is that we can
freely choose 4—1=3 values. Suppose we choose 27, 8, and 19 as the three values. Given
these three values and the information that the mean of the four values is 20, the fourth value
is 80 — 27 — 8 — 19 = 26. Thus, once we have chosen three values, the fourth value is auto-
matically determined. Consequently, the number of degrees of freedom for this example is
4f=n-l=4-l=3
We subtract 1 from n because we lose 1 degree of freedom to calculate the mean.
Table V of Appendix C lists the values of t for the given number of degrees of freedom and
areas in the right tail of a t distribution. Because the t distribution is symmetric, these are also the
values of — t for the same number of degrees of freedom and the same areas in the left tail of the
t distribution. Example 8^1- describes how to read Table V of Appendix C.
■ EXAMPLE 8-4
Find the value of t for 16 degrees of freedom and .05 area in the right tail of a t distribution curve.
Reading the t distribution table.
Solution In Table V of Appendix C, we locate 16 in the column of degrees of freedom
(labeled df) and .05 in the row of Area in the right tail under the t distribution curve at the
top of the table. The entry at the intersection of the row of 16 and the column of .05, which
is 1 .746, gives the required value of t. The relevant portion of Table V of Appendix C is shown
here as Table 8.2. The value of t read from the t distribution table is shown in Figure 8.6.
Table 8.2 Determining t for 16 df and .05 Area in the Right Tail
Area in the right tail
df
Area in the Right Tail
Under the t Distribution Curve
.10 gun —
.025 . . . .001
1
2
3
EH
75
3.078 6.314
1.886 2.920
1.638 2.353
1.337 |1.746|<-
1.293 1.665
1.282 1.645
12.706 . . . 318.309
4.303 . . . 22.327
3.182 . . . 10.215
2.120 . . . 3.686
1.992 . . . 3.202
1.960 . . . 3.090
The required value of t for 1 6 df and
.05 area in the right tail
Figure 8.6 The value of t for 16 df and .05
area in the right tail.
1 — .05
1.746
L
This is the required
value of t
8.4 Estimation of a Population Mean: <x Not Known 357
Because of the symmetric shape of the t distribution curve, the value of t for 16 degrees of
freedom and .05 area in the left tail is —1.746. Figure 8.7 illustrates this case.
Figure 8.7 The value of t for 16 df and .05 area in the
left tail.
-1 .746 t
8.4.2 Confidence Interval for fx Using the t Distribution
To reiterate, when the conditions mentioned under Cases I and II in the beginning of Section
8.4 hold true, we use the t distribution to construct a confidence interval for the population
mean, /x.
When the population standard deviation <x is not known, then we replace it by the sample
standard deviation s, which is its estimator. Consequently, for the standard deviation of I, we use
s
for = a I Vn . Note that the value of is a point estimate of Oj.
Confidence Interval for fx Using the t Distribution The (1 - a)100% confidence interval for i± is
where
s
S " = Vn
The value of t is obtained from the t distribution table for n — 1 degrees of freedom and the
given confidence level. Here tsj. is the margin of error of the estimate; that is,
E = tsj
Examples 8-5 and 8-6 describe the procedure of constructing a confidence interval for /jl
using the t distribution.
■ EXAMPLE 8-5
Dr. Moore wanted to estimate the mean cholesterol level for all adult men living in Hartford.
He took a sample of 25 adult men from Hartford and found that the mean cholesterol level
for this sample is 186 mg/dL with a standard deviation of 12 mg/dL. Assume that the cho-
lesterol levels for all adult men in Hartford are (approximately) normally distributed. Con-
struct a 95% confidence interval for the population mean /jl.
Solution Here, a is not known, n < 30, and the population is normally distributed. There-
fore, we will use the t distribution to make a confidence interval for /jl. From the given
information,
n = 25, x = 186, s = 12,
and Confidence level = 95%, or .95
The value of is
= 2.40
Constructing a 95%
confidence interval for fi
using the t distribution.
358 Chapter 8 Estimation of the Mean and Proportion
To find the value of t, we need to know the degrees of freedom and the area under the t dis-
tribution curve in each tail.
Degrees of freedom = n — 1=25 — 1=24
To find the area in each tail, we divide the confidence level by 2 and subtract the number ob-
tained from .5. Thus,
Area in each tail
(.95/2)
.4750 = .025
From the t distribution table, Table V of Appendix C, the value of t for df = 24 and .025 area
in the right tail is 2.064. The value of t is shown in Figure 8.8.
Figure 8.8 The value of t.
.025 — |
.4750
" N W/= 24
.4750 \.
1 — .025
-2.064
2.064
When we substitute all values in the formula for the confidence interval for /j,, the 95%
confidence interval is
jc ± % = 186 ± 2.064(2.40) = 186 ± 4.95 = 181.05 to 190.95
Thus, we can state with 95% confidence that the mean cholesterol level for all adult men
living in Hartford is between 181.05 and 190.95 mg/dL.
Note that x — 186 is a point estimate of j± in this example, and 4.95 is the margin of error.
Constructing a 99%
confidence interval for fi using
the t distribution.
■ EXAMPLE 8-6
Sixty-four randomly selected adults who buy books for general reading were asked how much
they usually spend on books per year. The sample produced a mean of $1450 and a standard
deviation of $300 for such annual expenses. Determine a 99% confidence interval for the cor-
responding population mean.
Solution From the given information,
n = 64, x = $1450, s = $300,
and Confidence level = 99%, or .99
Here a is not known, but the sample size is large (n > 30). Hence, we will use the t distri-
bution to make a confidence interval for /jl. First we calculate the standard deviation of x, the
number of degrees of freedom, and the area in each tail of the t distribution:
300
$37.50
' 1 Vn V64
df = n - 1 = 64 - 1 = 63
Area in each tail = .5 - (.99/2) = .5 - .4950
.005
From the t distribution table, t = 2.656 for 63 degrees of freedom and .005 area in the right
tail. The 99% confidence interval for /j, is
x ± tt s = $1450 ± 2.656(37.50)
= $1450 ± $99.60 = $1350.40 to $1549.60
Thus, we can state with 99% confidence that based on this sample the mean annual ex-
penditure on books by all adults who buy books for general reading is between $1350.40 and
$1549.60. ■
8.4 Estimation of a Population Mean: <x Not Known 359
Again, we can decrease the width of a confidence interval for yu, either by lowering the con-
fidence level or by increasing the sample size, as was done in Section 8.3. However, increasing
the sample size is the better alternative.
Note: What If the Sample Size Is Too Large?
In the above section, when cr is not known, we used the t distribution to make a confidence in-
terval for fx in Cases I and II. Note that in Case II, the sample size is large. If we have access to
technology, it does not matter how large (greater than 30) the sample size is, we can use the t
distribution. However, if we are using the t distribution table (Table V of Appendix C), this may
pose a problem. Usually such a table goes only up to a certain number of degrees of freedom.
For example, Table V in Appendix C goes only up to 75 degrees of freedom. Thus, if the sam-
ple size is larger than 76 in this section, we cannot use Table V to find the t value for the given
confidence level to use in the confidence interval in this section. In such a situation when n is
too large (for example, 500) and is not included in the t distribution table, there are two options:
1. Use the t value from the last row (the row of °°) in Table V.
2. Use the normal distribution as an approximation to the t distribution.
Note that the t values you will obtain from the last row of the t distribution table are the
same as obtained from the normal distribution table for the same confidence levels, the only
difference being the decimal places. To use the normal distribution as an approximation to the
t distribution to make a confidence interval for j±, the procedure is exactly like the one in Sec-
tion 8.3, except that now we replace <x by s, and by s^..
Again, note that here we can use the normal distribution as a convenience and as an ap-
proximation, but if we can, we should use the t distribution by using technology. Exercises 8.50,
8.51, and 8.65 at the end of this section present such situations.
EXERCISES
CONCEPTS AND PROCEDURES
8.36 Briefly explain the similarities and the differences between the standard normal distribution and the
t distribution.
8.37 What are the parameters of a normal distribution and a / distribution? Explain.
8.38 Briefly explain the meaning of the degrees of freedom for a t distribution. Give one example.
8.39 What assumptions must hold true to use the t distribution to make a confidence interval for fi7
8.40 Find the value of t for the t distribution for each of the following.
a. Area in the right tail = .05 and df = 12 b. Area in the left tail = .025 and n = 66
c. Area in the left tail = .001 and df = 49 d. Area in the right tail = .005 and n = 24
8.41 a. Find the value of t for the / distribution with a sample size of 21 and area in the left tail equal to .10.
b. Find the value of t for the t distribution with a sample size of 14 and area in the right tail equal
to .025.
c. Find the value of t for the t distribution with 45 degrees of freedom and .001 area in the right tail.
d. Find the value of t for the t distribution with 37 degrees of freedom and .005 area in the left tail.
8.42 For each of the following, find the area in the appropriate tail of the t distribution,
a. t = 2.467 and df = 28 b. t = - 1.672 and df = 58
c. t = -2.670 and n = 55 d. t = 2.383 and n = 23
8.43 For each of the following, find the area in the appropriate tail of the t distribution,
a. t = - 1.302 and df =42 b. t = 2.797 and n = 25
c. t = 1.397 and n = 9 d. t = -2.383 and df = 67
8.44 Find the value of t from the t distribution table for each of the following.
a. Confidence level = 99% and df = 13
b. Confidence level = 95% and n = 36
c. Confidence level = 90% and df = 16
360 Chapter 8 Estimation of the Mean and Proportion
8.45 a. Find the value of t from the t distribution table for a sample size of 22 and a confidence level of 95%.
b. Find the value of / from the / distribution table for 60 degrees of freedom and a 90% confidence
level.
c. Find the value of t from the t distribution table for a sample size of 24 and a confidence level of
99%.
8.46 A sample of 18 observations taken from a normally distributed population produced the following data.
28.4 27.3 25.5 25.5 31.1 23.0 26.3 24.6 28.4
37.2 23.9 28.7 27.9 25.1 27.2 25.3 22.6 22.7
a. What is the point estimate of /x?
b. Make a 99% confidence interval for /x.
c. What is the margin of error of estimate for /x in part b?
8.47 A sample of 11 observations taken from a normally distributed population produced the following
data.
-7.1 10.3 8.7 -3.6 -6.0 -7.5 5.2 3.7 9.8 -4.4 6.4
a. What is the point estimate of /x?
b. Make a 95% confidence interval for /x.
c. What is the margin of error of estimate for /x in part b?
8.48 Suppose, for a sample selected from a normally distributed population, x = 68.50 and s = 8.9.
a. Construct a 95% confidence interval for /x assuming = 16.
b. Construct a 90% confidence interval for /x assuming = 16. Is the width of the 90% confidence
interval smaller than the width of the 95% confidence interval calculated in part a? If yes, explain
why.
c. Find a 95% confidence interval for /x assuming n = 25. Is the width of the 95% confidence in-
terval for /x with n = 25 smaller than the width of the 95% confidence interval for xi with n = 16
calculated in part a? If so, why? Explain.
8.49 Suppose, for a sample selected from a population, x = 25.5 and s = 4.9.
a. Construct a 95% confidence interval for /x assuming n = 47.
b. Construct a 99% confidence interval for xi assuming n = 47. Is the width of the 99% confidence
interval larger than the width of the 95% confidence interval calculated in part a? If yes, explain
why.
c. Find a 95% confidence interval for /x assuming n = 32. Is the width of the 95% confidence in-
terval for fju with n = 32 larger than the width of the 95% confidence interval for /x with n = 47
calculated in part a? If so, why? Explain.
8.50 a. A sample of 100 observations taken from a population produced a sample mean equal to 55.32
and a standard deviation equal to 8.4. Make a 90% confidence interval for /x.
b. Another sample of 100 observations taken from the same population produced a sample mean
equal to 57.40 and a standard deviation equal to 7.5. Make a 90% confidence interval for /x.
c. A third sample of 100 observations taken from the same population produced a sample mean equal
to 56.25 and a standard deviation equal to 7.9. Make a 90% confidence interval for /x.
d. The true population mean for this population is 55.80. Which of the confidence intervals con-
structed in parts a through c cover this population mean and which do not?
8.51 a. A sample of 400 observations taken from a population produced a sample mean equal to 92.45
and a standard deviation equal to 12.20. Make a 98% confidence interval for /x.
b. Another sample of 400 observations taken from the same population produced a sample mean
equal to 91.75 and a standard deviation equal to 14.50. Make a 98% confidence interval for /x.
c. A third sample of 400 observations taken from the same population produced a sample mean equal
to 89.63 and a standard deviation equal to 13.40. Make a 98% confidence interval for xx.
d. The true population mean for this population is 90.65. Which of the confidence intervals con-
structed in parts a through c cover this population mean and which do not?
■ APPLICATIONS
8.52 A random sample of 16 airline passengers at the Bay City airport showed that the mean time spent
waiting in line to check in at the ticket counters was 31 minutes with a standard deviation of 7 minutes.
Construct a 99% confidence interval for the mean time spent waiting in line by all passengers at this air-
port. Assume that such waiting times for all passengers are normally distributed.
8.4 Estimation of a Population Mean: <x Not Known 361
8.53 A random sample of 20 acres gave a mean yield of wheat equal to 41.2 bushels per acre with a stan-
dard deviation of 3 bushels. Assuming that the yield of wheat per acre is normally distributed, construct
a 90% confidence interval for the population mean fi.
8.54 A May 8, 2008, report on National Public Radio (www.npr.org) noted that the average age of first-
time mothers in the United States is slightly higher than 25 years. Suppose that a random sample of 60
first-time mothers taken recently produced an average age of 25.9 years and a standard deviation of 3.2
years. Calculate a 90% confidence interval for the average age of all current first-time mothers.
8.55 Foods that have less than .50 gram of trans fat per serving are allowed to list the trans fat content as
zero grams on labels. A random sample of 32 foods that list the amount of trans fat as zero grams but con-
tain partially hydrogenated oils (the primary source of added trans fat) were evaluated for the amount of
trans fat per serving. The sample mean and standard deviation of trans fat were found to be .34 and .062
gram, respectively. Calculate a 95% confidence interval for the average amount of trans fat per serving in
all foods that are listed as having zero grams of trans fat per serving but contain partially hydrogenated oils.
8.56 The high cost of health care is a matter of major concern for a large number of families. A random
sample of 25 families selected from an area showed that they spend an average of $253 per month on
health care with a standard deviation of $47. Make a 98% confidence interval for the mean health care
expenditure per month incurred by all families in this area. Assume that the monthly health care expen-
ditures of all families in this area have a normal distribution.
8.57 Jack's Auto Insurance Company customers sometimes have to wait a long time to speak to a customer
service representative when they call regarding disputed claims. A random sample of 25 such calls yielded
a mean waiting time of 22 minutes with a standard deviation of 6 minutes. Construct a 99% confidence
interval for the population mean of such waiting times. Assume that such waiting times for the popula-
tion follow a normal distribution.
8.58 A random sample of 36 mid-sized cars tested for fuel consumption gave a mean of 26.4 miles per
gallon with a standard deviation of 2.3 miles per gallon.
a. Find a 99% confidence interval for the population mean, /x.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this inter-
val be reduced? Describe all possible alternatives. Which alternative is the best and why?
8.59 The mean time taken to design a house plan by 40 architects was found to be 23 hours with a stan-
dard deviation of 3.75 hours.
a. Construct a 98% confidence interval for the population mean /jl.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this inter-
val be reduced? Describe all possible alternatives. Which alternative is the best and why?
8.60 The following data give the speeds (in miles per hour), as measured by radar, of 10 cars traveling
on Interstate 1-15.
76 72 80 68 76 74 71 78 82 65
Assuming that the speeds of all cars traveling on this highway have a normal distribution, construct a 90%
confidence interval for the mean speed of all cars traveling on this highway.
8.61 A company randomly selected nine office employees and secretly monitored their computers for one
month. The times (in hours) spent by these employees using their computers for non-job-related activi-
ties (playing games, personal communications, etc.) during this month are given below.
7 1 29 8 1 14 1 41 6
Assuming that such times for all employees are normally distributed, make a 95% confidence interval for
the corresponding population mean for all employees of this company.
8.62 The following data are the times (in seconds) of eight finalists in the Girls' 100-meter dash at the
North Carolina 1A High School Track and Field Championships (Source: prepinsiders.blogspot.com).
12.25 12.37 12.68 12.84 12.90 12.97 13.02 13.35
Assume that these times represent a random sample of times for girls who would qualify for the finals of
this event, and that the population distribution of such times is normal. Determine the 98% confidence in-
terval for the average time in the Girls' 100-meter dash finals using these data.
8.63 Fifteen randomly selected ripe Macintosh apples had the following weights (in ounces).
8.9 6.8 7.2 8.3 8.1 7.9 7.1 8.0
8.5 6.7 7.0 7.4 7.7 6.2 9.2
Assume that the weight of a ripe Macintosh apple is normally distributed. Construct a 95% confidence in-
terval for the average weight of all ripe Macintosh apples.
362
Chapter 8 Estimation of the Mean and Proportion
8.64 According to liposuction4you.com, the maximum amount of fat and fluid that can be removed safely
during a liposuction procedure is 6 liters. Suppose that the following data represent the amount of fat and
fluid removed during 12 randomly selected liposuction procedures. Assume that the population distribu-
tion of such amounts is normal.
1.84 2.66 2.96 2.42 2.88 2.86
3.66 3.65 2.33 2.66 3.20 2.24
a. What is the point estimate of the corresponding population mean?
b. Construct a 98% confidence interval for the corresponding population mean.
8.65 An article in the Los Angeles Times (latimesblogs.latimes.com/pardonourdust/) quoted from the Na-
tional Association of Realtors that "we now sell our homes and move an average of every six years." Sup-
pose that the average time spent living in a house prior to selling it for a random sample of 400 recent
home sellers was 6.18 years and the sample standard deviation was 2.87 years.
a. What is the point estimate of the corresponding population mean?
b. Construct a 98% confidence interval for the average time spent living in a house prior to selling
it for all home owners. What is the margin of error for this estimate?
*8.66 You are working for a supermarket. The manager has asked you to estimate the mean time taken
by a cashier to serve customers at this supermarket. Briefly explain how you will conduct this study. Col-
lect data on the time taken by any supermarket cashier to serve 40 customers. Then estimate the popula-
tion mean. Choose your own confidence level.
*8.67 You are working for a bank. The bank manager wants to know the mean waiting time for all cus-
tomers who visit this bank. She has asked you to estimate this mean by taking a sample. Briefly explain
how you will conduct this study. Collect data on the waiting times for 45 customers who visit a bank.
Then estimate the population mean. Choose your own confidence level.
8.5 Estimation of a Population Proportion:
Large Samples
Often we want to estimate the population proportion or percentage. (Recall that a percentage is
obtained by multiplying the proportion by 100.) For example, the production manager of a com-
pany may want to estimate the proportion of defective items produced on a machine. A bank
manager may want to find the percentage of customers who are satisfied with the service pro-
vided by the bank.
Again, if we can conduct a census each time we want to find the value of a population pro-
portion, there is no need to learn the procedures discussed in this section. However, we usually
derive our results from sample surveys. Hence, to take into account the variability in the results
obtained from different sample surveys, we need to know the procedures for estimating a pop-
ulation proportion.
Recall from Chapter 7 that the population proportion is denoted by p, and the sample
proportion is denoted by p. This section explains how to estimate the population proportion, p,
using the sample proportion, p. The sample proportion, p, is a sample statistic, and it possesses
a sampling distribution. From Chapter 7, we know that for large samples:
1. The sampling distribution of the sample proportion, p, is (approximately) normal.
2. The mean, pup, of the sampling distribution of p is equal to the population proportion, p.
3. The standard deviation, cr-, of the sampling distribution of the sample proportion, p, is
\fpq/n, where q = 1 — p.
Remember ► i n the case of a proportion, a sample is considered to be large if tip and nq are both greater than 5.
If p and q are not known, then np and nq should each be greater than 5 for the sample to be large.
When estimating the value of a population proportion, we do not know the values of p and q.
Consequently, we cannot compute cr-. Therefore, in the estimation of a population proportion,
we use the value of s~ as an estimate of ap. The value of Sp is calculated using the following
formula.
8.5 Estimation of a Population Proportion: Large Samples 363
Estimator of the Standard Deviation of p The value of sp„ which gives a point estimate of o-p„ is
calculated as follows. Here, sp is an estimator of crp r
IM
p V n
The sample proportion, p, is the point estimator of the corresponding population propor-
tion, p. Then to find the confidence interval for p, we add to and subtract from p a number that
is called the margin of error, E.
Confidence Interval for the Population Proportion, p The (1 - a) 100% confidence interval for
the population proportion, p, is
p ± ZSp
The value of z used here is obtained from the standard normal distribution table for the given
confidence level, and Sp = vpq/n. The term z Sp is called the margin of error, E.
Examples 8-7 and 8-8 illustrate the procedure for constructing a confidence interval for p.
■ EXAMPLE 8-7
According to a survey conducted by Pew Research Center in June 2009, 44% of people aged
18 to 29 years said that religion is very important to them. Suppose this result is based on a
sample of 1000 people aged 18 to 29 years.
(a) What is the point estimate of the corresponding population proportion?
(b) Find, with a 99% confidence level, the percentage of all people aged 18 to 29 years
who will say that religion is very important to them. What is the margin of error of
this estimate?
Solution Let p be the proportion of all people aged 18 to 29 years who will say that reli-
gion is very important to them, and let p be the corresponding sample proportion. From the
given information,
n = 1000, p = .44, and q = 1 - p = 1 - .44 = .56
First, we calculate the value of the standard deviation of the sample proportion as follows:
^ = VT = V^ooo- = - 01569713
Note that np and riq are both greater than 5. (The reader should check this condition.) Con-
sequently, the sampling distribution of p is approximately normal, and we will use the normal
distribution to make a confidence interval about p.
(a) The point estimate of the proportion of all people aged 18 to 29 years who will say
that religion is very important to them is equal to .44; that is,
Point estimate of p = p = .44
(b) The confidence level is 99%, or .99. To find z for a 99% confidence level, first we find
the area in each of the two tails of the normal distribution curve, which is
(1 - .99)/2 = .0050. Then, we look for .0050 and .0050 + .99 = .9950 areas in the
normal distribution table to find the two values of z- These two z values are
Finding the point estimate and
99% confidence interval
for p: large sample.
364 Chapter 8 Estimation of the Mean and Proportion
(approximately) —2.58 and 2.58. Thus, we will use z = 2.58 in the confidence interval
formula. Substituting all the values in the confidence interval formula for p, we obtain
p ± zs p = .44 ± 2.58(.01569713) = .44 ± .04
= .40 to .48 or 40% to 48%
Thus, we can state with 99% confidence that .40 to .48, or 40% to 48%, of all
people aged 18 to 29 years will say that religion is very important to them.
The margin of error associated with this estimate of p is .04 or 4%, that is,
Margin of error = ±.04 or ±4% ■
EXAMPLE 8-8
According to a Harris Interactive survey of 2401 adults conducted in April 2009, 25% of adults
do not drink alcohol. Construct a 97% confidence interval for the corresponding population
proportion.
Solution Let p be the proportion of all adults who do not drink alcohol, and let p be the
corresponding sample proportion. From the given information,
n = 2401, p = .25, q = 1 - p = 1 - .25 = .75
and
Confidence level = 97% or .97
The standard deviation of the sample proportion is
interval for p: large sample.
-- .00883699
2401
From the normal distribution table, the value of z for the 97% confidence interval is 2.17. Note
that to find this z value, you will look for the areas .0150 and .9850 in Table IV. Substituting
all the values in the formula, we obtain the 97% confidence interval for p,
p ± zs ? = .25 ± 2.17(.00883699) = .25 ± .019
= .231 to .269, or 23.1% to 26.9%
Thus, we can state with 97% confidence that the proportion of all adults who do not drink
alcohol is between .231 and .269. This confidence interval can be converted into a percentage
interval as 23.1% to 26.9%. ■
Again, we can decrease the width of a confidence interval for p either by lowering the con-
fidence level or by increasing the sample size. However, lowering the confidence level is not a
good choice because it simply decreases the likelihood that the confidence interval contains p.
Hence, to decrease the width of a confidence interval for p, we should always increase the
sample size.
8.5.1 Determining the Sample Size
for the Estimation of Proportion
Just as we did with the mean, we can also determine the sample size for estimating the popu-
lation proportion, p. This sample size will yield an error of estimate that may not be larger than
a predetermined margin of error. By knowing the sample size that can give us the required re-
sults, we can save our scarce resources by not taking an unnecessarily large sample. From Sec-
tion 8.5, the margin of error, E, of the interval estimation of the population proportion is
E = zsp = z X
By manipulating this expression algebraically, we obtain the following formula to find the re-
quired sample size given E, p, q, and z.
USA TODAY Snapshots®
Cover your ears
Which sound is most
frustrating to hear?
Baby crying
21%
Dog barking
13%
Source: Baby Orajel su rvey or 1 ,004 ad u Irs
by Kelron Research. Margin of error
margin ±ja percent jge pomes
WHICH
SOUND
IS THE MOST
FRUSTRATING
TO HEAR?
By Michelle He.1ly.1ncl Sam Ward. USA TODAY
The above chart shows the percentage of adults who list different items as the most frustrating to hear. The
results are based on a survey of 1004 adults conducted by Kelton Research. As an example, according to
this poll, 39% of these adults said that a car alarm is the worst noise to hear, 28% mentioned a jackham-
mer, and so on. Using the procedure learned in this section, we can then make the confidence intervals
for the various proportions as shown in the table below. Note that the percentages in the chart add up to
101% because of the rounding.
Sound That is Most
Frustrating to Hear
Sample Proportion
Confidence Interval
Car alarm
.39
.39 ± ZSp
Jackhammer
.28
.28 ± zSp-
Baby crying
.21
.21 ± zSp
Dog barking
.13
.13 ± zs^
For each confidence interval listed in the table, we can substitute the value of z and the value of s^, which
is calculated as F° r example, suppose we want to find a 96% confidence interval for the proportion
of all adults who consider car alarm to be the worst noise to hear. This confidence interval is determined
as follows.
m /(.39)(.61) „,„„„ c
S S = \\ — = \\ — r^r; — = -01539325
p V n V 1004
p ± zs p - = .39 ± 2.05(.0 1539325) = .39 ± .03 = .36 to .42
Thus, we can expect 36% to 42% of all adults to list car alarm as the worst noise to hear.
We can find the confidence intervals for the population proportions of other three categories the same way.
Source: The chart reproduced with
permission from USA TODAY, August 20,
2009. Copyright © 2009, USA TODAY.
Determining the Sample Size for the Estimation of p Given the confidence level and the values
of p and q, the sample size that will produce a predetermined margin of error E of the confidence
interval estimate of p is
Z 2 pq
n = —
E 2
365
366 Chapter 8 Estimation of the Mean and Proportion
We can observe from this formula that to find re, we need to know the values of p and q.
However, the values of p and q are not known to us. In such a situation, we can choose one of
the following alternatives.
1. We make the most conservative estimate of the sample size n by using p = .50 and q = .50.
For a given E, these values of p and q will give us the largest sample size in comparison
to any other pair of values of p and q because the product of p = .50 and q = .50 is greater
than the product of any other pair of values for p and q.
2. We take a preliminary sample (of arbitrarily determined size) and calculate p and q for this
sample. Then, we use these values of p and q to find re.
Examples 8-9 and 8-10 illustrate how to determine the sample size that will produce the
error of estimation for the population proportion within a predetermined margin of error value.
Example 8-9 gives the most conservative estimate of re, and Example 8-10 uses the results from
a preliminary sample to determine the required sample size.
Determining the most
conservative estimate of n
for the estimation of p.
■ EXAMPLE 8-9
Lombard Electronics Company has just installed a new machine that makes a part that is
used in clocks. The company wants to estimate the proportion of these parts produced by
this machine that are defective. The company manager wants this estimate to be within .02
of the population proportion for a 95% confidence level. What is the most conservative es-
timate of the sample size that will limit the margin of error to within .02 of the population
proportion?
Solution The company manager wants the 95% confidence interval to be
p ± .02
Therefore,
E = .02
The value of z for a 95% confidence level is 1.96. For the most conservative estimate of the
sample size, we will use p = .50 and q = .50. Hence, the required sample size is
z 2 pq (1.96) 2 (.50)(.50)
n = — =— = ; — = 2401
E 1
(■02) 2
Thus, if the company takes a sample of 2401 parts, there is a 95% chance that the estimate
of p will be within .02 of the population proportion. ■
Determining n for the
estimation of p using
preliminary sample results.
■ EXAMPLE 8-10
Consider Example 8-9 again. Suppose a preliminary sample of 200 parts produced by this
machine showed that 7% of them are defective. How large a sample should the company select
so that the 95% confidence interval for p is within .02 of the population proportion?
Solution Again, the company wants the 95% confidence interval for p to be
p ± .02
Hence,
E = .02
The value of z for a 95% confidence level is 1.96. From the preliminary sample,
p = .07 and q = 1 - .07 = .93
Using these values of p and q, we obtain
= z 2 pg = (1.96) 2 (.07)(.93) = (3.8416)(.07)(.93)
n ~ E 2 (.02) 2
.0004
625.22 « 626
8.5 Estimation of a Population Proportion: Large Samples 367
Thus, if the company takes a sample of 626 items, there is a 95% chance that the estimate
of p will be within .02 of the population proportion. However, we should note that this sample
size will produce the margin of error within .02 only if p is .07 or less for the new sample. If
p for the new sample happens to be much higher than .07, the margin of error will not be
within .02. Therefore, to avoid such a situation, we may be more conservative and take a much
larger sample than 626 items. I
EXERCISES
CONCEPTS AND PROCEDURES
8.68 What assumption(s) must hold true to use the normal distribution to make a confidence interval for
the population proportion, pi
8.69 What is the point estimator of the population proportion, pi
8.70 Check if the sample size is large enough to use the normal distribution to make a confidence inter-
val for p for each of the following cases.
a. n = 50 and p = .25 b. n = 160 and p = .03
c. n = 400 and p = .65 d. n = 75 and p = .06
8.71 Check if the sample size is large enough to use the normal distribution to make a confidence interval
for p for each of the following cases.
a. n = 80 and p = .85 b. n = 110 and p = .98
c. n = 35 and p = .40 d. n = 200 and p = .08
8.72 a. A sample of 300 observations taken from a population produced a sample proportion of .63. Make
a 95% confidence interval for p.
b. Another sample of 300 observations taken from the same population produced a sample propor-
tion of .59. Make a 95% confidence interval for p.
c. A third sample of 300 observations taken from the same population produced a sample propor-
tion of .67. Make a 95% confidence interval for p.
d. The true population proportion for this population is .65. Which of the confidence intervals con-
structed in parts a through c cover this population proportion and which do not?
8.73 a. A sample of 1100 observations taken from a population produced a sample proportion of .32.
Make a 90% confidence interval for p.
b. Another sample of 1 100 observations taken from the same population produced a sample pro-
portion of .36. Make a 90% confidence interval for p.
c. A third sample of 1100 observations taken from the same population produced a sample propor-
tion of .30. Make a 90% confidence interval for p.
d. The true population proportion for this population is .34. Which of the confidence intervals con-
structed in parts a through c cover this population proportion and which do not?
8.74 A sample of 200 observations selected from a population produced a sample proportion equal to .91.
a. Make a 90% confidence interval for p.
b. Construct a 95% confidence interval for p.
c. Make a 99% confidence interval for p.
d. Does the width of the confidence intervals constructed in parts a through c increase as the confi-
dence level increases? If yes, explain why.
8.75 A sample of 200 observations selected from a population gave a sample proportion equal to .27.
a. Make a 99% confidence interval for p.
b. Construct a 97% confidence interval for p.
c. Make a 90% confidence interval for p.
d. Does the width of the confidence intervals constructed in parts a through c decrease as the con-
fidence level decreases? If yes, explain why.
8.76 A sample selected from a population gave a sample proportion equal to .73.
a. Make a 99% confidence interval for p assuming n = 100.
b. Construct a 99% confidence interval for p assuming n = 600.
c. Make a 99% confidence interval for p assuming n = 1500.
d. Does the width of the confidence intervals constructed in parts a through c decrease as the sam-
ple size increases? If yes, explain why.
368 Chapter 8 Estimation of the Mean and Proportion
8.77 A sample selected from a population gave a sample proportion equal to .31.
a. Make a 95% confidence interval forp assuming n = 1200.
b. Construct a 95% confidence interval for p assuming n = 500.
c. Make a 95% confidence interval for p assuming n = 80.
d. Does the width of the confidence intervals constructed in parts a through c increase as the sam-
ple size decreases? If yes, explain why.
8.78 a. How large a sample should be selected so that the margin of error of estimate for a 99% confi-
dence interval for p is .035 when the value of the sample proportion obtained from a preliminary
sample is .29?
b. Find the most conservative sample size that will produce the margin of error for a 99% confi-
dence interval for p equal to .035.
8.79 a. How large a sample should be selected so that the margin of error of estimate for a 98% confi-
dence interval for p is .045 when the value of the sample proportion obtained from a preliminary
sample is .53?
b. Find the most conservative sample size that will produce the margin of error for a 98% confi-
dence interval for p equal to .045.
8.80 Determine the most conservative sample size for the estimation of the population proportion for the
following.
a. E = .025, confidence level = 95%
b. E = .05, confidence level = 90%
c. E = .015, confidence level = 99%
8.81 Determine the sample size for the estimation of the population proportion for the following, where
p is the sample proportion based on a preliminary sample.
a. E = .025, p = .16, confidence level = 99%
b. E = .05, p = .85, confidence level = 95%
c. E = .015, p = .97, confidence level = 90%
■ APPLICATIONS
8.82 On November 15, 2006, carefair.com reported that 40% of women aged 30 years and older would
rather get Botox injections than spend a week in Paris. The survey consisted of 175 women in the speci-
fied age group.
a. What is the point estimate of the corresponding population proportion?
b. Construct a 98% confidence interval for the proportion of all women aged 30 years and older who
would rather get Botox injections than spend a week in Paris. What is the margin of error for this
estimate?
8.83 The express check-out lanes at Wally's Supermarket are limited to customers purchasing 12 or fewer
items. Cashiers at this supermarket have complained that many customers who use the express lanes have
more than 12 items. A recently taken random sample of 200 customers entering express lanes at this su-
permarket found that 74 of them had more than 12 items.
a. Construct a 98% confidence interval for the percentage of all customers at this supermarket who
enter express lanes with more than 12 items.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this inter-
val be reduced? Discuss all possible alternatives. Which alternative is the best?
8.84 According to an estimate, 11% of junk mail is thrown out without being opened (Source: www.
uoregon.edu/~recycle/events_topicsjunkmail_text.htm). Suppose that this percentage is based on a ran-
dom sample of 1400 pieces of junk mail.
a. What is the point estimate of the corresponding population proportion?
b. Construct a 95% confidence interval for the proportion of all pieces of junk mail that is thrown
out without being opened. What is the margin of error for this estimate?
8.85 It is said that happy and healthy workers are efficient and productive. A company that manu-
factures exercising machines wanted to know the percentage of large companies that provide on-site
health club facilities. A sample of 240 such companies showed that 96 of them provide such facilities
on site.
a. What is the point estimate of the percentage of all such companies that provide such facilities on site?
b. Construct a 97% confidence interval for the percentage of all such companies that provide such
facilities on site. What is the margin of error for this estimate?
8.5 Estimation of a Population Proportion: Large Samples 369
8.86 A mail-order company promises its customers that the products ordered will be mailed within 72
hours after an order is placed. The quality control department at the company checks from time to time
to see if this promise is fulfilled. Recently the quality control department took a sample of 50 orders and
found that 35 of them were mailed within 72 hours of the placement of the orders.
a. Construct a 98% confidence interval for the percentage of all orders that are mailed within 72
hours of their placement.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this interval
be reduced? Discuss all possible alternatives. Which alternative is the best?
8.87 In a random sample of 50 homeowners selected from a large suburban area, 19 said that they had
serious problems with excessive noise from their neighbors.
a. Make a 99% confidence interval for the percentage of all homeowners in this suburban area who
have such problems.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this inter-
val be reduced? Discuss all possible alternatives. Which option is best?
8.88 A jumbo mortgage is a mortgage with a loan amount above the industry-standard definition of con-
ventional conforming loan limits. As of January 2009, approximately 2.57% of people who took out a
jumbo mortgage during the previous 12 months were at least 60 days late on their payments. Suppose that
this percentage is based on a random sample of 1430 people who took out a jumbo mortgage during the
previous 12 months.
a. Construct a 95% confidence interval for the proportion of all people who took out a jumbo mort-
gage during the previous 12 months and were at least 60 days late on their payments.
b. Suppose the confidence interval obtained in part a is too wide. How can the width of this inter-
val be reduced? Discuss all possible alternatives. Which alternative is the best?
8.89 A Centers for Disease Control and Prevention survey about cell phone use noted that 14.7% of U.S.
households are wireless-only, which means that the household members use only cell phones and do not
have a landline. Suppose that this percentage is based on a random sample of 855 U.S. households (Source:
http://www.cdc.gov/nchs/data/nhsr/nhsr014.pdf).
a. Construct a 95% confidence interval for the proportion of all U.S. households that are wireless-only.
b. Explain why we need to construct a confidence interval. Why can we not simply say that 14.7%
of all U.S. households are wireless-onlyl
8.90 A researcher wanted to know the percentage of judges who are in favor of the death penalty. He took
a random sample of 15 judges and asked them whether or not they favor the death penalty. The responses
of these judges are given here.
Yes No Yes Yes No No No Yes
Yes No Yes Yes Yes No Yes
a. What is the point estimate of the population proportion?
b. Make a 95% confidence interval for the percentage of all judges who are in favor of the death penalty.
8.91 The management of a health insurance company wants to know the percentage of its policyholders
who have tried alternative treatments (such as acupuncture, herbal therapy, etc.). A random sample of
24 of the company's policyholders were asked whether or not they have ever tried such treatments. The
following are their responses.
Yes
No
No
Yes
No
Yes
No
No
No
Yes
No
No
Yes
No
Yes
No
No
No
Yes
No
No
No
Yes
No
a. What is the point estimate of the corresponding population proportion?
b. Construct a 99% confidence interval for the percentage of this company's policyholders who have
tried alternative treatments.
8.92 Tony's Pizza guarantees all pizza deliveries within 30 minutes of the placement of orders. An agency
wants to estimate the proportion of all pizzas delivered within 30 minutes by Tony's. What is the most
conservative estimate of the sample size that would limit the margin of error to within .02 of the popula-
tion proportion for a 99% confidence interval?
8.93 Refer to Exercise 8.92. Assume that a preliminary study has shown that 93% of all Tony's pizzas
are delivered within 30 minutes. How large should the sample size be so that the 99% confidence inter-
val for the population proportion has a margin of error of .02?
8.94 A consumer agency wants to estimate the proportion of all drivers who wear seat belts while driv-
ing. Assume that a preliminary study has shown that 76% of drivers wear seat belts while driving. How
370 Chapter 8 Estimation of the Mean and Proportion
large should the sample size be so that the 99% confidence interval for the population proportion has a
margin of error of .03?
8.95 Refer to Exercise 8.94. What is the most conservative estimate of the sample size that would limit
the margin of error to within .03 of the population proportion for a 99% confidence interval?
*8.96 You want to estimate the proportion of students at your college who hold off-campus (part-time or
full-time) jobs. Briefly explain how you will make such an estimate. Collect data from 40 students at your
college on whether or not they hold off-campus jobs. Then calculate the proportion of students in this sam-
ple who hold off-campus jobs. Using this information, estimate the population proportion. Select your own
confidence level.
*8.97 You want to estimate the percentage of students at your college or university who are satisfied with
the campus food services. Briefly explain how you will make such an estimate. Select a sample of 30 stu-
dents and ask them whether or not they are satisfied with the campus food services. Then calculate the
percentage of students in the sample who are satisfied. Using this information, find the confidence inter-
val for the corresponding population percentage. Select your own confidence level.
USES AND MISUSES... NATIONAL VERSUS LOCAL UNEMPLOYMENT RATE
Reading a newspaper article, you learn that the national unem-
ployment rate is 9.4%. The next month you read another article
that states that a recent survey in your area, based on a random
sample of the labor force, estimates that the local unemployment
rate is 9.0% with a margin of error of .5%. Thus, you conclude that
the unemployment rate in your area is somewhere between 8.5%
and 9.5%.
So, what does this say about the local unemployment picture
in your area versus the national unemployment situation? Since a
major portion of the interval for the local unemployment rate is
below 9.4%, is it reasonable to conclude that the local unemploy-
ment rate is below the national unemployment rate? Not really.
When looking at the confidence interval, you have some degree of
confidence, usually between 90% and 99%. If we use z = ±1.96
to calculate the margin of error, which is the z value for a 95%
confidence level, we can state that there is a 95% chance that the
local unemployment rate falls in the interval we obtain by using
the margin of error. However, since 9.4% is in the interval for the
local unemployment rate, the one thing that you can say is that it
appears reasonable to conclude that the local and national unem-
ployment rates are not different. However, if the national rate was
9.6%, then a conclusion that the two rates differ is reasonable
because we are confident that the local unemployment rate falls
between 8.5% and 9.5%.
When making conclusions based on the types of confidence in-
tervals you have learned and will learn in this course, you will only
be able to conclude that either there is a difference or there is not a
difference. However, the methods you will learn in Chapter 9 will also
allow you to determine the validity of a conclusion that states that
the local rate is lower (or higher) than the national rate.
Glossary
Confidence interval An interval constructed around the value of
a sample statistic to estimate the corresponding population parameter.
Confidence level Confidence level, denoted by (1 — cr)100%, that
states how much confidence we have that a confidence interval con-
tains the true population parameter.
Degrees of freedom (df ) The number of observations that can be
chosen freely. For the estimation of /jl using the t distribution, the
degrees of freedom are n — 1.
Estimate The value of a sample statistic that is used to find the
corresponding population parameter.
Estimation A procedure by which a numerical value or values are
assigned to a population parameter based on the information col-
lected from a sample.
Estimator The sample statistic that is used to estimate a popula-
tion parameter.
Interval estimate An interval constructed around the point esti-
mate that is likely to contain the corresponding population parame-
ter. Each interval estimate has a confidence level.
Margin of error The quantity that is subtracted from and added
to the value of a sample statistic to obtain a confidence interval for
the corresponding population parameter.
Point estimate The value of a sample statistic assigned to the cor-
responding population parameter.
t distribution A continuous distribution with a specific type of
bell-shaped curve with its mean equal to and standard deviation
equal to Vdf/(df - 2).
Supplementary Exercises
8.98 Because of inadequate public school budgets and lack of money available to teachers for class-
room materials, many teachers often use their own money to buy materials used in the classrooms. A
random sample of 100 public school teachers selected from an eastern state showed that they spent
an average of $273 on such materials during the 2009 school year. The population standard deviation
was $60.
a. What is the point estimate of the mean of such expenses incurred during the 2009 school year
by all public school teachers in this state?
b. Make a 95% confidence interval for the corresponding population mean.
8.99 A bank manager wants to know the mean amount owed on credit card accounts that become delin-
quent. A random sample of 100 delinquent credit card accounts taken by the manager produced a mean
amount owed on these accounts equal to $2640. The population standard deviation was $578.
a. What is the point estimate of the mean amount owed on all delinquent credit card accounts at
this bank?
b. Construct a 97% confidence interval for the mean amount owed on all delinquent credit card
accounts for this bank.
8.100 York Steel Corporation produces iron rings that are supplied to other companies. These rings
are supposed to have a diameter of 24 inches. The machine that makes these rings does not produce
each ring with a diameter of exactly 24 inches. The diameter of each of the rings varies slightly. It is
known that when the machine is working properly, the rings made on this machine have a mean diameter
of 24 inches. The standard deviation of the diameters of all rings produced on this machine is always
equal to .06 inch. The quality control department takes a sample of 25 such rings every week, calcu-
lates the mean of the diameters for these rings, and makes a 99% confidence interval for the popula-
tion mean. If either the lower limit of this confidence interval is less than 23.975 inches or the upper
limit of this confidence interval is greater than 24.025 inches, the machine is stopped and adjusted. A
recent such sample of 25 rings produced a mean diameter of 24.015 inches. Based on this sample, can
you conclude that the machine needs an adjustment? Explain. Assume that the population distribution
is normal.
8.101 Yunan Corporation produces bolts that are supplied to other companies. These bolts are supposed
to be 4 inches long. The machine that makes these bolts does not produce each bolt exactly 4 inches long.
It is known that when the machine is working properly, the mean length of the bolts made on this ma-
chine is 4 inches. The standard deviation of the lengths of all bolts produced on this machine is always
equal to .04 inch. The quality control department takes a sample of 20 such bolts every week, calculates
the mean length of these bolts, and makes a 98% confidence interval for the population mean. If either
the upper limit of this confidence interval is greater than 4.02 inches or the lower limit of this confidence
interval is less than 3.98 inches, the machine is stopped and adjusted. A recent such sample of 20 bolts
produced a mean length of 3.99 inches. Based on this sample, will you conclude that the machine needs
an adjustment? Assume that the population distribution is normal.
8.102 A hospital administration wants to estimate the mean time spent by patients waiting for treatment
at the emergency room. The waiting times (in minutes) recorded for a random sample of 35 such patients
are given below.
30
7
68
76
47
60
51
64
25
35
29
30
35
62
96
104
58
32
32
102
27
45
11
64
62
72
39
92
84
47
12
33
55
84
36
Construct a 99% confidence interval for the corresponding population mean. Use the t distribution.
8.103 A local gasoline dealership in a small town wants to estimate the average amount of gasoline that
people in that town use in a 1-week period. The dealer asked 44 randomly selected customers to keep a
diary of their gasoline usage, and this information produced the following data on gas used (in gallons)
by these people during a 1-week period.
23.1
13.6
25.8
10.0
7.6
18.9
26.6
23.8
12.3
15.8
21.0
26.9
22.9
18.3
23.5
21.6
15.5
23.5
11.8
15.3
11.9
19.2
14.5
9.6
12.1
18.0
20.6
14.2
7.1
13.2
5.3
13.1
10.9
10.5
5.1
5.2
6.5
8.3
10.5
7.4
7.4
5.3
10.6
13.0
372 Chapter 8 Estimation of the Mean and Proportion
Construct a 95% confidence interval for the average weekly gas usage by people in this town. Use the t
distribution.
8.104 A random sample of 25 life insurance policyholders showed that the average premium they pay on
their life insurance policies is $685 per year with a standard deviation of $74. Assuming that the life in-
surance policy premiums for all life insurance policyholders have a normal distribution, make a 99% con-
fidence interval for the population mean, /jl.
8.105 A drug that provides relief from headaches was tried on 18 randomly selected patients. The exper-
iment showed that the mean time to get relief from headaches for these patients after taking this drug was
24 minutes with a standard deviation of 4.5 minutes. Assuming that the time taken to get relief from a
headache after taking this drug is (approximately) normally distributed, determine a 95% confidence in-
terval for the mean relief time for this drug for all patients.
8.106 A survey of 500 randomly selected adult men showed that the mean time they spend per week
watching sports on television is 9.75 hours with a standard deviation of 2.2 hours. Construct a 90% con-
fidence interval for the population mean, /a.
8.107 A random sample of 300 female members of health clubs in Los Angeles showed that they
spend, on average, 4.5 hours per week doing physical exercise with a standard deviation of .75 hour.
Find a 98% confidence interval for the population mean.
8.108 A computer company that recently developed a new software product wanted to estimate the mean
time taken to learn how to use this software by people who are somewhat familiar with computers. A ran-
dom sample of 12 such persons was selected. The following data give the times taken (in hours) by these
persons to learn how to use this software.
1.75 2.25 2.40 1.90 1.50 2.75
2.15 2.25 1.80 2.20 3.25 2.60
Construct a 95% confidence interval for the population mean. Assume that the times taken by all persons
who are somewhat familiar with computers to learn how to use this software are approximately normally
distributed.
8.109 A company that produces 8-ounce low-fat yogurt cups wanted to estimate the mean number of
calories for such cups. A random sample of 10 such cups produced the following numbers of calories.
147 159 153 146 144 148 163 153 143 158
Construct a 99% confidence interval for the population mean. Assume that the numbers of calories for
such cups of yogurt produced by this company have an approximately normal distribution.
8.110 An insurance company selected a sample of 50 auto claims filed with it and investigated those
claims carefully. The company found that 12% of those claims were fraudulent.
a. What is the point estimate of the percentage of all auto claims filed with this company that
are fraudulent?
b. Make a 99% confidence interval for the percentage of all auto claims filed with this company
that are fraudulent.
8.111 A casino player has grown suspicious about a specific roulette wheel. Specifically, this player be-
lieves that the slots for the numbers and 00, which can lead to larger payoffs, are slightly smaller than
the rest of 36 slots, which means that the ball would land in these two slots less often than it would if all
of the slots were of the same size. This player watched 430 spins on this roulette wheel, and found that
the ball landed in or 00 slot 14 times.
a. What is the value of the point estimate of the proportion of all roulette spins on this wheel in
which the ball would land in or 00 slot?
b. Construct a 95% confidence interval for the proportion of all roulette spins on this wheel in
which the ball would land in or 00 slot.
c. If all of the slots on this wheel are of the same size, the ball should land in or 00 slot 5.26%
of the time. Based on the confidence interval you calculated in part b, does the player's suspi-
cion seem reasonable?
8.112 A sample of 20 managers was taken, and they were asked whether or not they usually take work
home. The responses of these managers are given below, where yes indicates they usually take work home
and no means they do not.
Yes Yes No No No Yes No No No No
Yes Yes No Yes Yes No No No No Yes
Make a 99% confidence interval for the percentage of all managers who take work home.
Supplementary Exercises 373
8.113 Salaried workers at a large corporation receive 2 weeks' paid vacation per year. Sixteen randomly
selected workers from this corporation were asked whether or not they would be willing to take a 3%
reduction in their annual salaries in return for 2 additional weeks of paid vacation. The following are the
responses of these workers.
No Yes No No Yes No No Yes
Yes No No No Yes No No No
Construct a 97% confidence interval for the percentage of all salaried workers at this corporation who
would accept a 3% pay cut in return for 2 additional weeks of paid vacation.
8.114 A researcher wants to determine a 99% confidence interval for the mean number of hours that adults
spend per week doing community service. How large a sample should the researcher select so that the
estimate is within 1.2 hours of the population mean? Assume that the standard deviation for time spent
per week doing community service by all adults is 3 hours.
8.115 An economist wants to find a 90% confidence interval for the mean sale price of houses in a state.
How large a sample should she select so that the estimate is within $3500 of the population mean? Assume
that the standard deviation for the sale prices of all houses in this state is $31,500.
8.116 A large city with chronic economic problems is considering legalizing casino gambling. The city
council wants to estimate the proportion of all adults in the city who favor legalized casino gambling.
What is the most conservative estimate of the sample size that would limit the margin of error to be within
.05 of the population proportion for a 95% confidence interval?
8.117 Refer to Exercise 8.116. Assume that a preliminary sample has shown that 63% of the adults in this
city favor legalized casino gambling. How large should the sample size be so that the 95% confidence
interval for the population proportion has a margin of error of .05?
Advanced Exercises
8.118 Let /jl be the hourly wage (excluding tips) for workers who provide hotel room service in a large
city. A random sample of a number (more than 30) of such workers yielded a 95% confidence interval for
/jl of $8.46 to $9.86 using the normal distribution with a known population standard deviation.
a. Find the value of x for this sample.
b. Find a 99% confidence interval for /x based on this sample.
8.119 In June 2008, SBRI Public Affairs conducted a telephone poll of 1004 adult Americans aged 18
and older. One of the questions asked was, "In the past year, was there ever a time when you . . . ?" Re-
spondents could choose more than one of the answers mentioned. Of the respondents, 64% said "cut back on
vacations or entertainment because of their cost," 37% said "failed to pay a bill on time," and 25% said "have
not gone to a doctor because of the cost." (Source: http://www.srbi.com/AmericansConcemEconomic.html.)
Using these results, find a 95% confidence interval for the corresponding population percentage for
each answer. Write a one-page report to present these results to a group of college students who have
not taken statistics. Your report should answer questions such as: (1) What is a confidence interval?
(2) Why is a range of values more informative than a single percentage? (3) What does 95% confi-
dence mean in this context? (4) What assumptions, if any, are you making when you construct each
confidence interval?
8.120 A group of veterinarians wants to test a new canine vaccine for Lyme disease. (Lyme disease is
transmitted by the bite of an infected deer tick.) In an area that has a high incidence of Lyme disease,
100 dogs are randomly selected (with their owners' permission) to receive the vaccine. Over a 12-month
period, these dogs are periodically examined by veterinarians for symptoms of Lyme disease. At the
end of 12 months, 10 of these 100 dogs are diagnosed with the disease. During the same 12-month pe-
riod, 18% of the unvaccinated dogs in the area have been found to have Lyme disease. Let p be the pro-
portion of all potential vaccinated dogs who would contract Lyme disease in this area.
a. Find a 95% confidence interval for p.
b. Does 18% lie within your confidence interval of part a? Does this suggest the vaccine might
or might not be effective to some degree?
c. Write a brief critique of this experiment, pointing out anything that may have distorted the
results or conclusions.
8.121 When one is attempting to determine the required sample size for estimating a population mean, and
the information on the population standard deviation is not available, it may be feasible to take a small pre-
liminary sample and use the sample standard deviation to estimate the required sample size, n. Suppose that
we want to estimate /j,, the mean commuting distance for students at a community college, to within 1 mile
with a confidence level of 95%. A random sample of 20 students yields a standard deviation of 4.1 miles. Use
374 Chapter 8 Estimation of the Mean and Proportion
this value of the sample standard deviation, s, to estimate the required sample size, n. Assume that the corre-
sponding population has a normal distribution.
8.122 A gas station attendant would like to estimate p, the proportion of all households that own more
than two vehicles. To obtain an estimate, the attendant decides to ask the next 200 gasoline customers how
many vehicles their households own. To obtain an estimate of p, the attendant counts the number of
customers who say there are more than two vehicles in their households and then divides this number by
200. How would you critique this estimation procedure? Is there anything wrong with this procedure that
would result in sampling and/or nonsampling errors? If so, can you suggest a procedure that would re-
duce this error?
8.123 A couple considering the purchase of a new home would like to estimate the average number of
cars that go past the location per day. The couple guesses that the number of cars passing this location per
day has a population standard deviation of 170.
a. On how many randomly selected days should the number of cars passing the location be
observed so that the couple can be 99% certain the estimate will be within 100 cars of the
true average?
b. Suppose the couple finds out that the population standard deviation of the number of cars
passing the location per day is not 170 but is actually 272. If they have already taken a sample
of the size computed in part a, what confidence does the couple have that their point estimate
is within 100 cars of the true average?
c. If the couple has already taken a sample of the size computed in part a and later finds out that
the population standard deviation of the number of cars passing the location per day is actually
130, they can be 99% confident their point estimate is within how much of the true average?
8.124 The U.S. Senate just passed a bill by a vote of 55-45 (with all 100 senators voting). A student who
took an elementary statistics course last semester says, "We can use these data to make a confidence in-
terval about p. We have n = 100 and p = 55/100 = .55." Hence, according to him, a 95% confidence
interval for p is
p ± zap = .55 ± 1.96 y 1QQ = .55 ± .098 = .452 to .648
Does this make sense? If not, what is wrong with the student's reasoning?
8.125 When calculating a confidence interval for the population mean p with a known population standard
deviation cr, describe the effects of the following two changes on the confidence interval: (1) doubling the
sample size, (2) quadrupling (multiplying by 4) the sample size. Give two reasons why this relationship
does not hold true if you are calculating a confidence interval for the population mean p with an unknown
population standard deviation.
8.126 At the end of Section 8.3, we noted that we always round up when calculating the minimum sam-
ple size for a confidence interval for p with a specified margin of error and confidence level. Using the
formula for the margin of error, explain why we must always round up in this situation.
8.127 Calculating a confidence interval for the proportion requires a minimum sample size. Calculate a
confidence interval, using any confidence level of 90% or higher, for the population proportion for each
of the following.
a. = 200 and p = .01 b. n= 160 and p = .9875
Explain why these confidence intervals reveal a problem when the conditions for using the normal ap-
proximation do not hold.
Self-Review Test
1. Complete the following sentences using the terms population parameter and sample statistic.
a. Estimation means assigning values to a based on the value of a .
b. An estimator is the used to estimate a .
c. The value of a is called the point estimate of the corresponding .
2. A 95% confidence interval for p can be interpreted to mean that if we take 100 samples of the same
size and construct 100 such confidence intervals for p, then
a. 95 of them will not include p b. 95 will include p c. 95 will include x
Self-Review Test
375
3.
The confidence level is denoted by
a. (1 - a) 100% b. 100a%
c. a
4.
The margin of error of the estimate for is
a. zoj (or fjj) b. cr/Vw (or s/Vrc)
c. o-j(orij)
5. Which of the following assumptions is not required to use the t distribution to make a confidence in-
terval for /x?
a. Either the population from which the sample is taken is (approximately) normally distributed or
n > 30.
b. The population standard deviation, o\ is not known.
c. The sample size is at least 10.
6. The parameter(s) of the t distribution is (are)
a. n b. degrees of freedom c. [l and degrees of freedom
7. A sample of 36 vacation homes built during the past 2 years in a coastal resort region gave a mean
construction cost of $159,000 with a population standard deviation of $27,000.
a. What is the point estimate of the corresponding population mean?
b. Make a 99% confidence interval for the mean construction cost for all vacation homes built in
this region during the past 2 years. What is the margin of error here?
8. A sample of 25 malpractice lawsuits filed against doctors showed that the mean compensation awarded
to the plaintiffs was $410,425 with a standard deviation of $74,820. Find a 95% confidence interval for
the mean compensation awarded to plaintiffs of all such lawsuits. Assume that the compensations awarded
to plaintiffs of all such lawsuits are normally distributed.
9. A poll on www.espn.com asked people who would they most want to play a round of golf with. The
choices mentioned were Michael Jordan, Ben Roethlisberger, Justin Timberlake, and Tiger Woods. Fifty-
five percent of the respondents chose Tiger Woods (33% chose Michael Jordan, and 6% chose Roethlis-
berger and Timberlake each). Although the sample is not a random sample, assume that the data came
from a random sample of 450 www.espn.com readers.
a. What is the value of the point estimate of the proportion of all www.espn.com readers who
would choose Tiger Woods when given the aforementioned set of choices?
b. Construct a 99% confidence interval for the proportion of all www.espn.com readers who would
choose Tiger Woods when given the aforementioned set of choices.
10. A company that makes toaster ovens has done extensive testing on the accuracy of its temperature-
setting mechanism. For a previous toaster model of this company, the standard deviation of the tempera-
tures when the mechanism is set for 350°F is 5.78°. Assume that this is the population standard deviation
for a new toaster model that uses the same temperature mechanism. How large a sample must be taken so
that the estimate of the mean temperature when the mechanism is set for 350°F is within 1.25° of the pop-
ulation mean temperature? Use a 95% confidence level.
11. A college registrar has received numerous complaints about the online registration procedure at her
college, alleging that the system is slow, confusing, and error prone. She wants to estimate the proportion
of all students at this college who are dissatisfied with the online registration procedure. What is the most
conservative estimate of the sample size that would limit the margin of error to be within .05 of the pop-
ulation proportion for a 90% confidence interval?
12. Refer to Problem 1 1 . Assume that a preliminary study has shown that 70% of the students surveyed
at this college are dissatisfied with the current online registration system. How large a sample should be
taken in this case so that the margin of error is within .05 of the population proportion for a 90% confi-
dence interval?
13. Dr. Garcia estimated the mean stress score before a statistics test for a random sample of 25 students.
She found the mean and standard deviation for this sample to be 7.1 (on a scale of 1 to 10) and 1.2, re-
spectively. She used a 97% confidence level. However, she thinks that the confidence interval is too wide.
How can she reduce the width of the confidence interval? Describe all possible alternatives. Which alter-
native do you think is best and why?
*14. You want to estimate the mean number of hours that students at your college work per week. Briefly
explain how you will conduct this study using a small sample. Take a sample of 12 students from your
college who hold a job. Collect data on the number of hours that these students spent working last week.
Then estimate the population mean. Choose your own confidence level. What assumptions will you make
to estimate this population mean?
376 Chapter 8 Estimation of the Mean and Proportion
*15. You want to estimate the proportion of people who are happy with their current jobs. Briefly ex-
plain how you will conduct this study. Take a sample of 35 persons and collect data on whether or not
they are happy with their current jobs. Then estimate the population proportion. Choose your own con-
fidence level.
Mini-Projects
■ MINI-PROJECT 8-1
A study performed by the Oregon Employment Agency and Bureau of Labor Statistics asked high school
students the age-old question, "What do you want to be when you grow up?" The study summary divided
the responses into different categories that are listed in the following table.
Category
Number of
Responses
Percentage of
Total
Professional and Related
94
47
Health Care
50
25
Management, Business, and Financial
22
11
Service
14
7
Sales and Related
12
6
Construction and Extraction
4
2
Other
4
2
Source: http://www.qualityinfo.org/olmisj/ArticleReader ?itemid=00006008.
Although the article noted that the study was not a scientific study, for the purposes of answers to the
question asked, we will perform the analysis as if the results came from a random sample selected from
all high school students in Oregon. Explain how you could use the data in the table to construct confi-
dence intervals for the true proportions of the different career categories. Include a discussion about any
issues with regard to sample size.
■ MINI-PROJECT 8-2
Consider the data set on the heights of NBA players that accompanies this text.
a. Take a random sample of 15 players, and find a 95% confidence interval for fi. Assume that the
heights of these players are normally distributed.
b. Repeat part a for samples of size 31 and 45, respectively.
c. Compare the widths of your three confidence intervals.
d. Now calculate the mean, /a, of the heights of all players. Do all of your confidence intervals con-
tain this ixl If not, which ones do not contain /x?
■ MINI-PROJECT 8-3
Here is a project that can involve a social activity and also show you the importance of making sure
that the underlying requirements are met prior to calculating a confidence interval. Invite some of your
friends over and buy a big bag of Milk Chocolate M&Ms. Take at least 40 random samples of 10
M&Ms each from the bag. Note that taking many random samples will reduce the risk of obtaining
some extremely odd results. Before eating the candy, calculate the proportion of brown candies for
each sample. Then, using each sample proportion, compute a 95% confidence interval for the propor-
tion of brown candies in all M&Ms. According to the company, the population proportion is .13, that
is, 13% of all M&Ms are brown. Determine what percentage of the confidence intervals contains the
population proportion .13. Is this percentage close to 95%? What happens if you increase your sam-
ple size to 20, and then to 50? If you want, you can use technology to simulate those random samples,
Decide for Yourself 377
which makes the process much faster. Besides, the candy will probably be eaten by the time you get
ready to take larger samples.
MINI-PROJECT 8-4
The following snapshot, reproduced from USA
TODAY of August 1 1 , 2009 (with permission of
USA TODAY), shows the results of a Randstad
survey in which 3032 adults were asked about
their desire to obtain a managerial position at
work.
As shown in the chart, 42% of adults aged
18-34 years stated that they do not want to be
managers.
a. Using the results of the study, calculate
a 95% confidence interval for the pro-
portion of all adults aged 1 8 to 34 years
who do not want to be managers. As-
sume that the survey included 758 adults
aged 18 to 34 years.
b. Take a random sample of 50 college stu-
dents within the specified age group (18
to 34 years) and ask them the same ques-
tion. Using your results, calculate a 95%
confidence interval for the proportion of
all adults aged 18-34 years who do not
USA TODAY Snapshots®
0^--^ Do you want to
r
t become a manager?
Percentage who said no:
\
Age 18-34
42%
Age 35-44
Age 45- 54
47%
50%
Age 55 and older
68%
Source: Randsiad survey of 3,032 adulrs 18 and older.
Weighted tu nrpment .niu.il puuuldlluii
By Jae Yang and Sam Ward. USA TODAY
C.
d.
want to be managers. If the sample size is not large enough to use the normal approximation, in-
crease the sample size to 75.
Compare the confidence intervals calculated in parts a and b. Are the results consistent between
the two surveys?
Is there any reason to believe that your results are not representative of the population of interest,
which is all adults aged 18 to 34 years? Explain why.
DECIDE FOR YOURSELF
Deciding About the Viability
of Poll Results
In the Decide for Yourself feature of Chapter 7, we discussed the
idea that underlies the procedures that are used to make projections
on election day. Here we discuss the process of collecting data in exit
polls.
Instead of selecting a simple random sample and choosing peo-
ple at random from a list of all voters (imagine the almost impossi-
ble process of preparing such a list), exit polls use what is called a
multistage sampling procedure. In this sampling technique, the first
stage involves randomly selecting a few voter precincts. If the U.S.
presidential election were based solely on the popular vote, these
precincts could be selected from a list of all precincts in the United
States. However, the presidential election is based on the Electoral
College system. Hence, the polling agencies need to select precincts
from each state in order to make sure that they have a sufficient sam-
ple size from each state. Then, interviewers who are stationed at each
selected precinct interview every kth voter, where k is dependent on
the expected number of voters at that precinct. Dr. Christian Potholm
of Bowdoin College in Brunswick, Maine, cited a problem during the
2004 presidential election. The following excerpt is taken from the
Web site http://www.bowdoin.edu/news/archives/lacademicnews/
001613. shtml that contained an interview with Dr. Potholm.
According to Dr. Potholm:
Those exit polls were really a disservice to polling. All
across the country these early polls took a bad sample
and exaggerated its impact. Whether it was done mali-
ciously or not, it was just bad polling not to balance
your sample. . . . What I think happened in the national
polls was that the initial polls were 58 percent women.
1. Explain why taking a poll early in the morning could produce
misleading results.
2. As mentioned in the above statement, women made up the major-
ity of the voters in those polls. Was this more likely to overrepresent
Kerry voters or Bush voters? Why?
3. Discuss some other potential issues with a time bias, which hap-
pens if a poll is taken at a specific time of the day.
378 Chapter 8 Estimation of the Mean and Proportion
ti
ECHNOLOGY
INSTRUCTION
Confidence Intervals for Population Means and Proportions
Z Interval
Inpt:Data EUEUE
<t:2
x: 11
n:65
C-Level: .951
Calculate
Screen 8.1
1. To find a confidence interval for a population mean i± given the population standard devia-
tion cr, select STAT >TESTS >ZInterval. If you have the data stored in a list, select
Data and enter the name of the list. If you have the summary statistics, choose Stats and
enter the sample mean and size. Enter your value for cr and the confidence level as a deci-
mal as C-Level. Select Calculate. (See Screen 8.1.)
2. To find a confidence interval for a population mean fi without knowing the population
standard deviation cr, select STAT >TESTS >TInterval. If you have the data stored in a
list, select Data and enter the name of the list. If you have the summary statistics, choose
Stats and enter the sample mean, standard deviation, and size. Enter your confidence level
as a decimal as C-Level. Select Calculate.
3. To find a confidence interval for a population proportion p, select STAT >TESTS >
1-PropZInt. Enter the number of successes as x and the sample size as n. Enter the
confidence level as a decimal as C-Level. Select Calculate.
1 -Sample Z [Test and Confidence Interval)
Samples in columns:
(* Summarized data
Sample size: |65
Mean: |ll
Standard deviation: 2
Test mean: \
(required for test)
Select
Graphs.
Help
OK
Options.
Cancel
1. To find a confidence interval for the population
mean /jl when the population standard deviation cr
is known, select Stat >Basic Statistics >1-Sam-
ple Z. If you have data on a variable entered in a
column of a Minitab spreadsheet, enter the name
of that column in the Samples in columns: box.
If you know the summary statistics, click next to
Summarized data and enter the values of the
Sample size and Mean in their respective boxes.
In both cases, enter the value of the population
standard deviation in the Standard deviation box.
(See Screen 8.2.) Click the Options button and
enter the Confidence level. Now click OK in both
windows. The confidence interval will appear in
the Session window.
Screen 8.2
2. To find a confidence interval for the population
mean /j, when the population standard deviation
cr is not known, select Stat >Basic Statistics
>1-Sample t. If you have data on a variable
entered in a column of a Minitab spreadsheet,
enter the name of that column in the Samples in
columns: box. If you know the summary statistics, click next to Summarized data and enter
the values of the Sample size, Mean, and Sample standard deviation in their respective
boxes. Click the Options button and enter the Confidence level. Now click OK in both win-
dows. The confidence interval will appear in the Session window.
3. To find a confidence interval for a population proportion p, select Stat >Basic Statistics
> 1-Proportion. If you have sample data (consisting of two values for success and failure)
entered in a column, select Samples in columns and type your column name in the box.
If, instead, you have the number of successes and the number of trials, select Summarized
data and enter them. Click the Options button and enter the Confidence level. Click OK
in both boxes. The confidence interval will appear in the Session window.
Technology Instruction 379
Descriptive Statistics
Input
Input Range:
Grouped By:
!■) Columns
<_■' Rows
1 1 Labels in first row
OK
Cancel
Help
Output options
© Output Range:
O New Worksheet Ply:
New Workbook
[^1 Summary statistics
[^1 Confidence Level for Mean
□ Kth Largest:
□ Kth Smallest:
$C|I
95
%
c
Column!
Mean
34.4
Standard Error
9_443634
Median
25.5
Mode
Standard Delation
29.86339
Sample Variance
891.8222.
KurtosiE
0.898782
Skevhiiess
1.171464
Range
93
Minimum^
5!
Maximum
98
Sum
344
Count
10
Confidence Level(95.D%)
21.36298
Screen 8.3
Screen 8.4
1. To calculate the margin of error for a confidence interval for a population mean when
the population standard deviation is unknown and the individual data values are avail-
able, first use the instructions to obtain the summary statistics (mean and standard
deviation) using the Analysis ToolPak presented in the Technology Instruction section
of Chapter 3. Then use the following additional step. After filling out all of the relevant
information in the Descriptive Statistics dialog box, check the box Confidence Level
for Mean and enter the confidence level as a percentage. Click OK. (See Screens
8.3 and 8.4.)
2. To find the margin of error for a confidence interval for a population mean when the pop-
ulation standard deviation a is known and the sample size n and the confidence level
1 - a are provided, type = CONFIDENCES, a, n). Note that for a 95% confidence
level, a = 1 - .95 = .05. (See Screens 8.5 and 8.6.)
1
A
B
Mean
11
2
Std. Dev.
~l
Size
65
4
Alpha
0.05
5
6
Margin of error
=C OXFEOEXCE(0.05^ 65)
Screen 8.5
1
B
1
Mean
11
2
Std. Dev.
~%|
Size
65
4
.Alpha
0.05
5
6
Margin of error
0.4S620":::!
Screen 8.6
380 Chapter 8 Estimation of the Mean and Proportion
TECHNOLOGY ASSIGNMENTS
TA8.1 The following data give the annual incomes (in thousands of dollars) before taxes for a sample
of 36 randomly selected families from a city.
21.6
33.0
25.6
37.9
50.0
148.1
50.1
21.5
70.0
72.8
58.2
85.4
91.2
57.0
72.2
45.0
95.0
27.8
92.8
79.4
45.3
76.0
48.6
69.3
40.6
69.0
75.5
57.5
49.7
75.1
96.3
44.5
84.0
43.0
61.7
126.0
Construct a 99% confidence interval for /jl assuming that the population standard deviation is $23.75 thousand.
TA8.2 The following data give the checking account balances (in dollars) on a certain day for a ran-
domly selected sample of 30 households.
500
100
650
1917
2200
500
180
3000
1500
1300
319
1500
1102
405
124
1000
134
2000
150
800
200
750
300
2300
40
1200
500
900
20
160
Construct a 97% confidence interval for /jl assuming that the population standard deviation is unknown.
TA8.3 Refer to Data Set I that accompanies this text on the prices of various products in different cities
across the country. Using the data on the cost of going to the dentist's office, make a 98% confidence
interval for the population mean /jl.
TA8.4 Refer to the Manchester Road Race data set that accompanies this text for all participants. Take
a sample of 100 observations from this data set.
a. Using the sample data, make a 95% confidence interval for the mean time taken to complete this race
by all participants.
b. Now calculate the mean time taken to run this race by all participants. Does the confidence interval
made in part a include this population mean?
TA8.5 Repeat Technology Assignment TA8.4 for a sample of 25 observations. Assume that the
distribution of times taken to run this race by all participants is approximately normal.
TA8.6 The following data give the prices (in thousands of dollars) of 16 recently sold houses in an area.
341 163 327 204 197 203 313 279
456 228 383 289 533 399 271 381
Construct a 99% confidence interval for the mean price of all houses in this area. Assume that the distri-
bution of prices of all houses in the given area is normal.
TA8.7 A researcher wanted to estimate the mean contributions made to charitable causes by major com-
panies. A random sample of 18 companies produced the following data on contributions (in millions of
dollars) made by them.
1.8 .6 1.2 .3 2.6 1.9 3.4 2.6 .2
2.4 1.4 2.5 3.1 .9 1.2 2.0 .8 1.1
Make a 98% confidence interval for the mean contributions made to charitable causes by all major com-
panies. Assume that the contributions made to charitable causes by all major companies have a normal
distribution.
TA8.8 A mail-order company promises its customers that their orders will be processed and mailed within
72 hours after an order is placed. The quality control department at the company checks from time to time
to see if this promise is kept. Recently the quality control department took a sample of 200 orders and
found that 176 of them were processed and mailed within 72 hours of the placement of the orders. Make
a 98% confidence interval for the corresponding population proportion.
TA8.9 One of the major problems faced by department stores is a high percentage of returns. The man-
ager of a department store wanted to estimate the percentage of all sales that result in returns. A sample
of 500 sales showed that 95 of them had products returned within the time allowed for returns. Make a
99% confidence interval for the corresponding population proportion.
TA8.10 One of the major problems faced by auto insurance companies is the filing of fraudulent claims.
An insurance company carefully investigated 1000 auto claims filed with it and found 108 of them to be
fraudulent. Make a 96% confidence interval for the corresponding population proportion.
Chapter
Hypothesis Tests About the Mean
and Proportion
When we travel by plane, does seat assignment make any difference? Do we always prefer a
specific seat when we fly? We sure do. In a survey of adults, 61% said that they prefer a window
seat, while 38% prefer an aisle seat. Only 1% of the respondents indicated a preference for the
middle seat. (See Case Study 9-2).
This chapter introduces the second topic in inferential statistics: tests of hypotheses. In a test of hypoth-
esis, we test a certain given theory or belief about a population parameter. We may want to find out,
using some sample information, whether or not a given claim (or statement) about a population param-
eter is true. This chapter discusses how to make such tests of hypotheses about the population mean,
IX, and the population proportion, p.
As an example, a soft-drink company may claim that, on average, its cans contain 12 ounces of
soda. A government agency may want to test whether or not such cans do contain, on average, 12
ounces of soda. As another example, according to the Giving USA Foundation, 75% of the total char-
itable contributions in 2008 were given by individuals. An economist may want to check if this
percentage is still true for this year. In the first of these two examples we are to test a hypothesis
about the population mean, /jl, and in the second example we are to test a hypothesis about the
population proportion, p.
9.1
9.2
Hypothesis Tests: An
Introduction
Hypothesis Tests About
/jl'. <r Known
Case Study 9-1 How
Crashes Affect Auto
Premiums
9.3 Hypothesis Tests About
fi: a Not Known
9.4 Hypothesis Tests About
a Population
Proportion: Large
Samples
Case Study 9-2 Favorite
Seat in the Plane
381
382 Chapter 9 Hypothesis Tests About the Mean and Proportion
9.1 Hypothesis Tests: An Introduction
Why do we need to perform a test of hypothesis? Reconsider the example about soft-drink
cans. Suppose we take a sample of 100 cans of the soft drink under investigation. We then
find out that the mean amount of soda in these 100 cans is 11.89 ounces. Based on this re-
sult, can we state that, on average, all such cans contain less than 12 ounces of soda and that
the company is lying to the public? Not until we perform a test of hypothesis can we make
such an accusation. The reason is that the mean, x = 11.89 ounces, is obtained from a sam-
ple. The difference between 12 ounces (the required average amount for the population) and
11.89 ounces (the observed average amount for the sample) may have occurred only because
of the sampling error (assuming that no nonsampling errors have been committed). Another
sample of 100 cans may give us a mean of 12.04 ounces. Therefore, we perform a test of
hypothesis to find out how large the difference between 12 ounces and 11.89 ounces is and
to investigate whether or not this difference has occurred as a result of chance alone. Now, if
11.89 ounces is the mean for all cans and not for just 100 cans, then we do not need to make
a test of hypothesis. Instead, we can immediately state that the mean amount of soda in all
such cans is less than 12 ounces. We perform a test of hypothesis only when we are making
a decision about a population parameter based on the value of a sample statistic.
9.1.1 Two Hypotheses
Consider as a nonstatistical example a person who has been indicted for committing a crime
and is being tried in a court. Based on the available evidence, the judge or jury will make one
of two possible decisions:
1. The person is not guilty.
2. The person is guilty.
At the outset of the trial, the person is presumed not guilty. The prosecutor's efforts are to prove
that the person has committed the crime and, hence, is guilty.
In statistics, the person is not guilty is called the null hypothesis and the person is guilty
is called the alternative hypothesis. The null hypothesis is denoted by H , and the alterna-
tive hypothesis is denoted by H x . In the beginning of the trial it is assumed that the person is
not guilty. The null hypothesis is usually the hypothesis that is assumed to be true to begin
with. The two hypotheses for the court case are written as follows (notice the colon after H
and //,):
Null hypothesis: H Q : The person is not guilty
Alternative hypothesis: H x : The person is guilty
In a statistics example, the null hypothesis states that a given claim (or statement) about
a population parameter is true. Reconsider the example of the soft-drink company's claim
that, on average, its cans contain 12 ounces of soda. In reality, this claim may or may not be
true. However, we will initially assume that the company's claim is true (that is, the company
is not guilty of cheating and lying). To test the claim of the soft-drink company, the null hy-
pothesis will be that the company's claim is true. Let p be the mean amount of soda in all
cans. The company's claim will be true if p = 12 ounces. Thus, the null hypothesis will be
written as
H : fi, = 12 ounces (The company's claim is true)
In this example, the null hypothesis can also be written as /jl > 12 ounces because the claim of
the company will still be true if the cans contain, on average, more than 12 ounces of soda. The
company will be accused of cheating the public only if the cans contain, on average, less than
12 ounces of soda. However, it will not affect the test whether we use an = or a > sign in the
null hypothesis as long as the alternative hypothesis has a < sign. Remember that in the null
hypothesis (and in the alternative hypothesis also) we use the population parameter (such as p.
or p) and not the sample statistic (such as x or p).
9.1 Hypothesis Tests: An Introduction
Definition
Null Hypothesis A null hypothesis is a claim (or statement) about a population parameter that
is assumed to be true until it is declared false.
The alternative hypothesis in our statistics example will be that the company's claim is false
and its soft-drink cans contain, on average, less than 12 ounces of soda — that is, /jl < 12 ounces.
The alternative hypothesis will be written as
H\. j± < 12 ounces (The company's claim is false)
Definition
Alternative Hypothesis An alternative hypothesis is a claim about a population parameter that
will be true if the null hypothesis is false.
Let us return to the example of the court trial. The trial begins with the assumption that
the null hypothesis is true — that is, the person is not guilty. The prosecutor assembles all the
possible evidence and presents it in the court to prove that the null hypothesis is false and
the alternative hypothesis is true (that is, the person is guilty). In the case of our statistics
example, the information obtained from a sample will be used as evidence to decide whether
or not the claim of the company is true. In the court case, the decision made by the judge
(or jury) depends on the amount of evidence presented by the prosecutor. At the end of
the trial, the judge (or jury) will consider whether or not the evidence presented by the pros-
ecutor is sufficient to declare the person guilty. The amount of evidence that will be consid-
ered to be sufficient to declare the person guilty depends on the discretion of the judge (or
jury).
9.1.2 Rejection and Nonrejection Regions
In Figure 9.1, which represents the court case, the point marked indicates that there is no
evidence against the person being tried. The farther we move toward the right on the hori-
zontal axis, the more convincing the evidence is that the person has committed the crime. We
have arbitrarily marked a point C on the horizontal axis. Let us assume that a judge (or jury)
considers any amount of evidence to the right of point C to be sufficient and any amount of
evidence to the left of C to be insufficient to declare the person guilty. Point C is called the
critical value or critical point in statistics. If the amount of evidence presented by the pros-
ecutor falls in the area to the left of point C, the verdict will reflect that there is not enough
evidence to declare the person guilty. Consequently, the accused person will be declared not
guilty. In statistics, this decision is stated as do not reject H . It is equivalent to saying that
there is not enough evidence to declare the null hypothesis false. The area to the left of point
C is called the nonrejection region; that is, this is the region where the null hypothesis is not
Not enough evidence to
declare the person guilty and,
hence, the null hypothesis is
not rejected in this region.
-Nonrejection region-
Enough evidence to declare
the person guilty and,
hence, the null hypothesis
is rejected in this region.
■ Rejection region -
C
t
Critical point
Figure 9.1 Nonrejection and rejection
regions for the court case.
Level of
evidence
384 Chapter 9 Hypothesis Tests About the Mean and Proportion
rejected. However, if the amount of evidence falls in the area to the right of point C, the ver-
dict will be that there is sufficient evidence to declare the person guilty. In statistics, this de-
cision is stated as reject H or the null hypothesis is false. Rejecting H is equivalent to say-
ing that the alternative hypothesis is true. The area to the right of point C is called the rejection
region; that is, this is the region where the null hypothesis is rejected.
9.1.3 Two Types of Errors
We all know that a court's verdict is not always correct. If a person is declared guilty at the end
of a trial, there are two possibilities.
1. The person has not committed the crime but is declared guilty (because of what may be
false evidence).
2. The person has committed the crime and is rightfully declared guilty.
In the first case, the court has made an error by punishing an innocent person. In statistics, this
kind of error is called a Type I or an a {alpha) error. In the second case, because the guilty
person has been punished, the court has made the correct decision. The second row in the shaded
portion of Table 9.1 shows these two cases. The two columns of Table 9.1, corresponding to
the person is not guilty and the person is guilty, give the two actual situations. Which one of
these is true is known only to the person being tried. The two rows in this table, corresponding
to the person is not guilty and the person is guilty, show the two possible court decisions.
Table 9.1 Four Possible Outcomes for a Court Case
Actual Situation
The Person Is
The Person
Not Guilty
Is Guilty
The person is
Correct
Type II or
Court's
not guilty
decision
/3 error
decision
The person
Type I or
Correct
is guilty
a error
decision
In our statistics example, a Type I error will occur when H Q is actually true (that is, the cans
do contain, on average, 12 ounces of soda), but it just happens that we draw a sample with a
mean that is much less than 12 ounces and we wrongfully reject the null hypothesis, H . The
value of a, called the significance level of the test, represents the probability of making a
Type I error. In other words, a is the probability of rejecting the null hypothesis, H , when in
fact it is true.
Definition
Type I Error A Type I error occurs when a true null hypothesis is rejected. The value of a
represents the probability of committing this type of error; that is,
a = P(H is rejected | H is true)
The value of a represents the significance level of the test.
The size of the rejection region in a statistics problem of a test of hypothesis depends on
the value assigned to a. In one approach to a test of hypothesis, we assign a value to a before
making the test. Although any value can be assigned to a, the commonly used values of a are
.01, .025, .05, and .10. Usually the value assigned to a does not exceed .10 (or 10%).
Now, suppose that in the court trial case the person is declared not guilty at the end of the
trial. Such a verdict does not indicate that the person has indeed not committed the crime. It is
9.1 Hypothesis Tests: An Introduction
385
possible that the person is guilty but there is not enough evidence to prove the guilt. Conse-
quently, in this situation there are again two possibilities.
1. The person has not committed the crime and is declared not guilty.
2. The person has committed the crime but, because of the lack of enough evidence, is de-
clared not guilty.
In the first case, the court's decision is correct. In the second case, however, the court has
committed an error by setting a guilty person free. In statistics, this type of error is called a
Type II or a f} (the Greek letter beta) error. These two cases are shown in the first row of the
shaded portion of Table 9.1.
In our statistics example, a Type II error will occur when the null hypothesis, H , is actu-
ally false (that is, the soda contained in all cans, on average, is less than 12 ounces), but it hap-
pens by chance that we draw a sample with a mean that is close to or greater than 12 ounces
and we wrongfully conclude do not reject H . The value of /3 represents the probability of mak-
ing a Type II error. It represents the probability that H is not rejected when actually H is false.
The value of 1 — /3 is called the power of the test. It represents the probability of not making
a Type II error.
Definition
Type II Error A Type II error occurs when a false null hypothesis is not rejected. The value of
/3 represents the probability of committing a Type II error; that is,
[} = P(H is not rejected | H is false)
The value of 1 — /3 is called the power of the test. It represents the probability of not making a
Type II error.
The two types of errors that occur in tests of hypotheses depend on each other. We cannot
lower the values of a and /3 simultaneously for a test of hypothesis for a fixed sample size.
Lowering the value of a will raise the value of /3, and lowering the value of /3 will raise the
value of a. However, we can decrease both a and /3 simultaneously by increasing the sam-
ple size. The explanation of how a and /3 are related and the computation of j3 are not within
the scope of this text.
Table 9.2, which is similar to Table 9.1, is written for the statistics problem of a test of hy-
pothesis. In Table 9.2 the person is not guilty is replaced by H is true, the person is guilty by
H is false, and the court's decision by decision.
Table 9.2 Four Possible Outcomes for a Test of Hypothesis
Actual Situation
H Is True
H Is False
Decision
Do not reject H
Correct
decision
Type II or
f} error
Reject H
Type I or
a error
Correct
decision
9.1.4 Tails of a Test
The statistical hypothesis-testing procedure is similar to the trial of a person in court but with
two major differences. The first major difference is that in a statistical test of hypothesis, the
partition of the total region into rejection and nonrejection regions is not arbitrary. Instead, it
depends on the value assigned to a (Type I error). As mentioned earlier, a is also called the
significance level of the test.
386 Chapter 9 Hypothesis Tests About the Mean and Proportion
The second major difference relates to the rejection region. In the court case, the rejection
region is on the right side of the critical point, as shown in Figure 9. 1 . However, in statistics,
the rejection region for a hypothesis-testing problem can be on both sides, with the nonrejec-
tion region in the middle, or it can be on the left side or right side of the nonrejection region.
These possibilities are explained in the next three parts of this section. A test with two rejec-
tion regions is called a two-tailed test, and a test with one rejection region is called a one-tailed
test. The one-tailed test is called a left-tailed test if the rejection region is in the left tail of the
distribution curve, and it is called a right-tailed test if the rejection region is in the right tail
of the distribution curve.
Definition
Tails of the Test A two-tailed test has rejection regions in both tails, a left-tailed test has the
rejection region in the left tail, and a right-tailed test has the rejection region in the right tail of
the distribution curve.
A Two-Tailed Test
According to a survey by Consumer Reports magazine conducted in 2008, a sample of sixth-
graders selected from New York schools showed that their backpacks weighed an average of
18.4 pounds (USA TODAY, August 3, 2009). Another magazine wants to check whether or not
this mean has changed since that survey. The key word here is changed. The mean weight of
backpacks for sixth-graders in New York has changed if it has either increased or decreased
since 2008. This is an example of a two-tailed test. Let /jl be the weight of backpacks for the
current sixth-graders in New York. The two possible decisions are
1. The mean weight of backpacks for sixth-graders in New York has not changed since 2008,
that is, currently /jl = 18.4 pounds.
2. The mean weight of backpacks for sixth-graders in New York has changed since 2008, that
is, currently /a + 18.4 pounds.
We write the null and alternative hypotheses for this test as follows:
H : /jl = 18.4 pounds (The mean weight of backpacks for sixth-graders in New York has not
changed)
H { : jx ¥= 18.4 pounds (The mean weight of backpacks for sixth-graders in New York has changed)
Whether a test is two-tailed or one-tailed is determined by the sign in the alternative hy-
pothesis. If the alternative hypothesis has a not equal to (¥=) sign, as in this example, it is a
two-tailed test. As shown in Figure 9.2, a two-tailed test has two rejection regions, one in each
tail of the distribution curve. Figure 9.2 shows the sampling distribution of x, assuming it has
a normal distribution. Assuming H is true, x has a normal distribution with its mean equal to
18.4 pounds (the value of /jl in H ). In Figure 9.2, the area of each of the two rejection regions
Figure 9.2 A two-tailed test.
This shaded
area is a/2
Rejection
region
H= 18.4
-Nonrejection region -
This shaded
area is a/2
Rejection
region
_ These two values are _
called the critical values
9.1 Hypothesis Tests: An Introduction 387
is a/2 and the total area of both rejection regions is a (the significance level). As shown in
this figure, a two-tailed test of hypothesis has two critical values that separate the two rejec-
tion regions from the nonrejection region. We will reject H if the value of x obtained from
the sample falls in either of the two rejection regions. We will not reject H if the value of x
lies in the nonrejection region. By rejecting H , we are saying that the difference between the
value of jju stated in H and the value of x obtained from the sample is too large to have oc-
curred because of the sampling error alone. Consequently, this difference is real. By not re-
jecting H , we are saying that the difference between the value of /jl stated in H and the value
of x obtained from the sample is small and it may have occurred because of the sampling er-
ror alone.
A Left-Tailed Test
Reconsider the example of the mean amount of soda in all soft-drink cans produced by a com-
pany. The company claims that these cans, on average, contain 12 ounces of soda. However, if
these cans contain less than the claimed amount of soda, then the company can be accused of
cheating. Suppose a consumer agency wants to test whether the mean amount of soda per can is
less than 12 ounces. Note that the key phrase this time is less than, which indicates a left-tailed
test. Let ijl be the mean amount of soda in all cans. The two possible decisions are as follows:
1. The mean amount of soda in all cans is equal to 12 ounces, that is, /jl = 12 ounces.
2. The mean amount of soda in all cans is less than 12 ounces, that is, jj, < 12 ounces.
The null and alternative hypotheses for this test are written as
H : ijl — 12 ounces (The mean is equal to 12 ounces)
H x : jx < 12 ounces (The mean is less than 12 ounces)
In this case, we can also write the null hypothesis as H : ijl > 12. This will not affect the re-
sult of the test as long as the sign in H l is less than (<).
When the alternative hypothesis has a less than (<) sign, as in this case, the test is always
left-tailed. In a left-tailed test, the rejection region is in the left tail of the distribution curve,
as shown in Figure 9.3, and the area of this rejection region is equal to a (the significance
level). We can observe from this figure that there is only one critical value in a left-tailed test.
Assuming H Q is true, the sampling distribution of x has a mean equal to 12 ounces (the
value of jx in H ). We will reject H if the value of x obtained from the sample falls in the re-
jection region; we will not reject H otherwise.
A Right-Tailed Test
To illustrate the third case, according to www.city-data.com, the average price of homes in West
Orange, New Jersey, was $461,216 in 2007. Suppose a real estate researcher wants to check
whether the current mean price of homes in this town is higher than $461,216. The key phrase
388 Chapter 9 Hypothesis Tests About the Mean and Proportion
in this case is higher than, which indicates a right-tailed test. Let /jl be the current mean price
of homes in this town. The two possible decisions are as follows:
1. The current mean price of homes in this town is not higher than $461,216, that is, currently
jx = $461,216.
2. The current mean price of homes in this town is higher than $461,216, that is, currently
> $461,216.
We write the null and alternative hypotheses for this test as follows:
H : /x = $461,216 (The current mean price of homes in this town is not higher than $461,216)
H x : ijl > $461,216 (The current mean price of homes in this area is higher than $461,216)
Note that here we can also write the null hypothesis as H : /x < $461,216, which states that the
current mean price of homes in this area is either equal to or less than $461,216. Again, the re-
sult of the test will not be affected by whether we use an equal to (=) or a less than or equal
to (s) sign in H as long as the alternative hypothesis has a greater than (>) sign.
When the alternative hypothesis has a greater than (>) sign, the test is always right-tailed.
As shown in Figure 9.4, in a right-tailed test, the rejection region is in the right tail of the dis-
tribution curve. The area of this rejection region is equal to a, the significance level. Like a left-
tailed test, a right-tailed test has only one critical value.
Figure 9.4 A right-tailed test.
Shaded
area is a
p = $461 ,21 6
Nonrejection region -
Rejection
region
Critical value
C
Again, assuming H is true, the sampling distribution of x has a mean equal to $461,216
(the value of fi in H ). We will reject H if the value of x obtained from the sample falls in the
rejection region. Otherwise, we will not reject H .
Table 9.3 summarizes the foregoing discussion about the relationship between the signs in
H a and H l and the tails of a test.
Table 9.3 Signs in H and H x and Tails of a Test
Two-Tailed
Left-Tailed
Right-Tailed
Test
Test
Test
Sign in the null
hypothesis H
= or >
= or <
Sign in the alternative
hypothesis H x
+
<
>
Rejection region
In both tails
In the left
tail
In the right
tail
Note that the null hypothesis always has an equal to (=) or a greater than or equal to (£)
or a less than or equal to (<) sign, and the alternative hypothesis always has a not equal to (#)
or a less than (<) or a greater than (>) sign.
9.1 Hypothesis Tests: An Introduction 389
In this text we will use the following two procedures to make tests of hypothesis.
1. The /7-value approach. Under this procedure, we calculate what is called the p-value for
the observed value of the sample statistic. If we have a predetermined significance level,
then we compare the /7-value with this significance level and make a decision. Note that
here p stands for probability.
2. The critical-value approach. In this approach, we find the critical value(s) from a table
(such as the normal distribution table or the t distribution table) and find the value of the
test statistic for the observed value of the sample statistic. Then we compare these two val-
ues and make a decision.
Remember, the procedures to be learned in this chapter assume that the sample taken
is a simple random sample.
EXERCISES
CONCEPTS AND PROCEDURES
9.1 Briefly explain the meaning of each of the following terms.
a. Null hypothesis b. Alternative hypothesis c. Critical point(s)
d. Significance level e. Nonrejection region f. Rejection region
g. Tails of a test h. Two types of errors
9.2 What are the four possible outcomes for a test of hypothesis? Show these outcomes by writing a table.
Briefly describe the Type I and Type II errors.
9.3 Explain how the tails of a test depend on the sign in the alternative hypothesis. Describe the signs in
the null and alternative hypotheses for a two-tailed, a left-tailed, and a right-tailed test, respectively.
9.4 Explain which of the following is a two-tailed test, a left-tailed test, or a right-tailed test.
a. H : p = 45, H x : p > 45 b. H : p = 23, H x : p + 23 c. H : p > 75, H t : p < 75
Show the rejection and nonrejection regions for each of these cases by drawing a sampling distribution
curve for the sample mean, assuming that it is normally distributed.
9.5 Explain which of the following is a two-tailed test, a left-tailed test, or a right-tailed test.
a. p = 12, Hi: p < 12 b. H : p < 85, p > 85 c. H : p = 33, H t : p + 33
Show the rejection and nonrejection regions for each of these cases by drawing a sampling distribution
curve for the sample mean, assuming that it is normally distributed.
9.6 Which of the two hypotheses (null and alternative) is initially assumed to be true in a test of hypothesis?
9.7 Consider H : p = 20 versus H t : p < 20.
a. What type of error would you make if the null hypothesis is actually false and you fail to reject it?
b. What type of error would you make if the null hypothesis is actually true and you reject it?
9.8 Consider H : p = 55 versus H t : p + 55.
a. What type of error would you make if the null hypothesis is actually false and you fail to reject it?
b. What type of error would you make if the null hypothesis is actually true and you reject it?
■ APPLICATIONS
9.9 Write the null and alternative hypotheses for each of the following examples. Determine if each is a
case of a two-tailed, a left-tailed, or a right-tailed test.
a. To test if the mean number of hours spent working per week by college students who hold jobs
is different from 20 hours
b. To test whether or not a bank's ATM is out of service for an average of more than 10 hours per month
c. To test if the mean length of experience of airport security guards is different from 3 years
d. To test if the mean credit card debt of college seniors is less than $1000
e. To test if the mean time a customer has to wait on the phone to speak to a representative of a mail-
order company about unsatisfactory service is more than 12 minutes
9.10 Write the null and alternative hypotheses for each of the following examples. Determine if each is
a case of a two-tailed, a left-tailed, or a right-tailed test.
a. To test if the mean amount of time spent per week watching sports on television by all adult men
is different from 9.5 hours
b. To test if the mean amount of money spent by all customers at a supermarket is less than $105
390
Chapter 9 Hypothesis Tests About the Mean and Proportion
c. To test whether the mean starting salary of college graduates is higher than $39,000 per year
d. To test if the mean waiting time at the drive-through window at a fast food restaurant during rush
hour differs from 10 minutes
e. To test if the mean hours spent per week on house chores by all housewives is less than 30
9.2 Hypothesis Tests About \x\ a Known
This section explains how to perform a test of hypothesis for the population mean /j, when the
population standard deviation cr is known. As in Section 8.3 of Chapter 8, here also there are
three possible cases as follows.
Case I. If the following three conditions are fulfilled:
1. The population standard deviation cr is known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is normally distributed,
then we use the normal distribution to perform a test of hypothesis about j± because from Sec-
tion 7.4.1 of Chapter 7 the sampling distribution of x is normal with its mean equal to fi and
the standard deviation equal to cr Y = cr/Vn, assuming that nIN ^ .05.
Case II. If the following two conditions are fulfilled:
1. The population standard deviation cr is known
2. The sample size is large (i.e., n > 30),
then, again, we use the normal distribution to perform a test of hypothesis about /jl because from
Section 7.4.2 of Chapter 7, due to the central limit theorem, the sampling distribution of x is
(approximately) normal with its mean equal to /jl and the standard deviation equal to cr T = o7 Vn,
assuming that n/N s .05.
Case III. If the following three conditions are fulfilled:
1. The population standard deviation cr is known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is not normally distributed (or the
shape of its distribution is unknown),
then we use a nonparametric method to perform a test of hypothesis about /jl.
This section will cover the first two cases. The procedure for performing a test of hypoth-
esis about i± is the same in both these cases. Note that in Case I, the population does not have
to be exactly normally distributed. As long as it is close to the normal distribution without any
outliers, we can use the normal distribution procedure. In Case II, although 30 is considered a
large sample, if the population distribution is very different from the normal distribution, then
30 may not be a large enough sample size for the sampling distribution of x to be normal and,
hence, to use the normal distribution.
The following chart summarizes the above three cases.
<T Is Known
I
Case I
1. n < 30
2. Population is normal
y
Case II
n > 30
/
Casein
1. n < 30
2. Population is not normal
Use the normal distribution
to test a hypothesis about /j.
Use a nonparametric method
to test a hypothesis about /j,
9.2 Hypothesis Tests About fj.: a Known 391
Below we explain two procedures, the p-value approach and the critical-value approach, to test
hypotheses about p, under Cases I and II. We will use the normal distribution to perform such tests.
Note that the two approaches — the p-value approach and the critical-value approach — are not
mutually exclusive. We do not need to use one or the other. We can use both at the same time.
1 . The p-Value Approach
In this procedure, we find a probability value such that a given null hypothesis is rejected for
any a (significance level) greater than this value and it is not rejected for any a less than this
value. The probability-value approach, more commonly called the p-value approach, gives
such a value. In this approach, we calculate the p-value for the test, which is defined as the
smallest level of significance at which the given null hypothesis is rejected. Using this p-value,
we state the decision. If we have a predetermined value of a, then we compare the value of p
with a and make a decision.
Definition
p-Value Assuming that the null hypothesis is true, the p-value can be defined as the probability that
a sample statistic (such as the sample mean) is at least as far away from the hypothesized value in
the direction of the alternative hypothesis as the one obtained from the sample data under consider-
ation. Note that the p-value is the smallest significance level at which the null hypothesis is rejected.
Using the p-value approach, we reject the null hypothesis if
p-value < a or a > p-value
and we do not reject the null hypothesis if
p-value > a or a < p-value
For a one-tailed test, the p-value is given by the area in the tail of the sampling distribu-
tion curve beyond the observed value of the sample statistic. Figure 9.5 shows the p-value for
a right-tailed test about p.. For a left-tailed test, the p-value will be the area in the lower tail of
the sampling distribution curve to the left of the observed value of 3c.
Figure 9.5 The p-value for a right-tailed test.
Value of x observed -
from the sample
For a two-tailed test, the p-value is twice the area in the tail of the sampling distribution
curve beyond the observed value of the sample statistic. Figure 9.6 shows the p-value for a two-
tailed test. Each of the areas in the two tails gives one-half the p-value.
-The sum of these two-
areas gives the p-value
Figure 9.6 The p-value for a two-tailed test.
Value of x observed -
from the sample
392 Chapter 9 Hypothesis Tests About the Mean and Proportion
To find the area under the normal distribution curve beyond the sample mean x, we first
find the z value for x using the following formula.
Calculating the z Value for x When using the normal distribution, the value of z for x for a test
of hypothesis about /jl is computed as follows:
— jju cr
where cr s = — -=
o- Tx V«
The value of z calculated for x using this formula is also called the observed value of z.
Then we find the area under the tail of the normal distribution curve beyond this value of
z. This area gives the p-value or one-half the p-value, depending on whether it is a one-tailed
test or a two-tailed test.
A test of hypothesis procedure that uses the /7-value approach involves the following
four steps.
Steps to Perform a Test of Hypothesis Using the p-Value Approach
1. State the null and alternative hypothesis.
2. Select the distribution to use.
3. Calculate the p-value.
4. Make a decision.
Examples 9-1 and 9-2 illustrate the calculation and use of the p-value to test a hypothesis
using the normal distribution.
Performing a hypothesis test
using the p-value approach
for a two-tailed test with the
normal distribution.
■ EXAMPLE 9-1
At Canon Food Corporation, it used to take an average of 90 minutes for new workers to learn
a food processing job. Recently the company installed a new food processing machine. The
supervisor at the company wants to find if the mean time taken by new workers to learn the
food processing procedure on this new machine is different from 90 minutes. A sample of 20
workers showed that it took, on average, 85 minutes for them to learn the food processing pro-
cedure on the new machine. It is known that the learning times for all new workers are nor-
mally distributed with a population standard deviation of 7 minutes. Find the /9-value for the
test that the mean learning time for the food processing procedure on the new machine is dif-
ferent from 90 minutes. What will your conclusion be if a = .01?
Solution Let /x be the mean time (in minutes) taken to learn the food processing procedure
on the new machine by all workers, and let x be the corresponding sample mean. From the
given information,
n = 20, x = 85 minutes, a = 7 minutes, and a = .01
To calculate the 79-value and perform the test, we apply the following four steps.
Step 1. State the null and alternative hypotheses.
H : /x — 90 minutes
Hi. fi 90 minutes
Note that the null hypothesis states that the mean time for learning the food processing pro-
cedure on the new machine is 90 minutes, and the alternative hypothesis states that this time
is different from 90 minutes.
9.2 Hypothesis Tests About fj.: a Known 393
Step 2. Select the distribution to use.
Here, the population standard deviation a is known, the sample size is small (n < 30), but
the population distribution is normal. Hence, the sampling distribution of x is normal with its
mean equal to /x and the standard deviation equal to cr T = a/Vn. Consequently, we will use
the normal distribution to find the p-value and make the test.
Step 3. Calculate the p-value.
The # sign in the alternative hypothesis indicates that the test is two-tailed. The p-value is
equal to twice the area in the tail of the sampling distribution curve of x to the left of 3c = 85, as
shown in Figure 9.7. To find this area, we first find the z value for x = 85 as follows:
a 7
cr, = — — = — — = 1 .56524758 minutes
Vn V20
W = 85-90
oi 1.56524758
The area to the left of x = 85 is equal to the area under the standard normal curve to the left
of z = —3.19. From the normal distribution table, the area to the left of z = —3.19 is .0007.
Consequently, the p-value is
p-value = 2(.0007) = .0014
Step 4. Make a decision.
Thus, based on the p-value of .0014, we can state that for any a (significance level) greater
than .0014 we will reject the null hypothesis stated in Step 1, and for any a less than or equal
to .0014, we will not reject the null hypothesis.
Because a = .01 is greater than the p-value of .0014, we reject the null hypothesis at this
significance level. Therefore, we conclude that the mean time for learning the food process-
ing procedure on the new machine is different from 90 minutes.
EXAMPLE 9-2
The management of Priority Health Club claims that its members lose an average of 10 pounds
or more within the first month after joining the club. A consumer agency that wanted to check
this claim took a random sample of 36 members of this health club and found that they lost
an average of 9.2 pounds within the first month of membership. The population standard de-
viation is known to be 2.4 pounds. Find the p-value for this test. What will your decision be
if a = .01? What if a = .05?
Performing a hypothesis test
using the p-value approach
for a one-tailed test with the
normal distribution.
Solution Let /x be the mean weight lost during the first month of membership by all mem-
bers of this health club, and let x be the corresponding mean for the sample. From the given
information,
n = 36, x = 9.2 pounds, and cr = 2.4 pounds
394 Chapter 9 Hypothesis Tests About the Mean and Proportion
The claim of the club is that its members lose, on average, 10 pounds or more within the first
month of membership. To perform the test using the p-value approach, we apply the follow-
ing four steps.
Step 1. State the null and alternative hypotheses.
H : /jl — 10 (The mean weight lost is 10 pounds or more.)
H t : fi < 10 (The mean weight lost is less than 10 pounds.)
Step 2. Select the distribution to use.
Here, the population standard deviation a is known, and the sample size is large (n > 30).
Hence, the sampling distribution of x is normal (due to the Central Limit Theorem) with its
mean equal to i± and the standard deviation equal to <r- K = a/Vn. Consequently, we will use
the normal distribution to find the p-value and perform the test.
Step 3. Calculate the p-value.
The < sign in the alternative hypothesis indicates that the test is left-tailed. The
p-value is given by the area to the left of x = 9.2 under the sampling distribution curve of x,
as shown in Figure 9.8. To find this area, we first find the z value for x = 9.2 as follows:
a 2.4
.40
36
_M _ 9.2 - 10
E " .40
-2.00
The area to the left of x = 9.2 under the sampling distribution of x is equal to the area under
the standard normal curve to the left of z = —2.00. From the normal distribution table, the area
to the left of z — —2.00 is .0228. Consequently,
p-value = .0228
Step 4. Make a decision.
Thus, based on the p-value of .0228, we can state that for any a (significance level) greater
than .0228 we will reject the null hypothesis stated in Step 1, and for any a less than or equal
to .0228 we will not reject the null hypothesis.
Since a = .01 is less than the p-value of .0228, we do not reject the null hypothesis at this
significance level. Consequently, we conclude that the mean weight lost within the first month
of membership by the members of this club is 10 pounds or more.
Now, because a = .05 is greater than the p-value of .0228, we reject the null hypothesis at
this significance level. Therefore, we conclude that the mean weight lost within the first month
of membership by the members of this club is less than 10 pounds.
2. The Critical-Value Approach
This is also called the traditional or classical approach. In this procedure, we have a predeter-
mined value of the significance level a. The value of a gives the total area of the rejection
9.2 Hypothesis Tests About fj.: a Known 395
region(s). First we find the critical value(s) of z from the normal distribution table for the given
significance level. Then we find the value of the test statistic z for the observed value of the
sample statistic x. Finally we compare these two values and make a decision. Remember, if the
test is one-tailed, there is only one critical value of z, and it is obtained by using the value of
a, which gives the area in the left or right tail of the normal distribution curve depending on
whether the test is left-tailed or right-tailed, respectively. However, if the test is two-tailed, there
are two critical values of z and they are obtained by using a/2 area in each tail of the normal
distribution curve. The value of the test statistic is obtained as follows.
Test Statistic In tests of hypotheses about fi using the normal distribution, the random variable
X — fJL (J
z = where <j t = — ;=
(Tx Vn
is called the test statistic. The test statistic can be defined as a rule or criterion that is used to
make the decision on whether or not to reject the null hypothesis.
A test of hypothesis procedure that uses the critical-value approach involves the following
five steps.
Steps to Perform a Test of Hypothesis with the Critical-Value Approach
1. State the null and alternative hypotheses.
2. Select the distribution to use.
3. Determine the rejection and nonrejection regions.
4. Calculate the value of the test statistic.
5. Make a decision.
Examples 9-3 and 9-4 illustrate the use of these five steps to perform tests of hypotheses
about the population mean jx. Example 9-3 is concerned with a two-tailed test, and Example
9-4 describes a one-tailed test.
■ EXAMPLE 9-3
The TIV Telephone Company provides long-distance telephone service in an area. According
to the company's records, the average length of all long-distance calls placed through this com-
pany in 2009 was 12.44 minutes. The company's management wanted to check if the mean
length of the current long-distance calls is different from 12.44 minutes. A sample of 150 such
calls placed through this company produced a mean length of 13.71 minutes. The standard de-
viation of all such calls is 2.65 minutes. Using the 2% significance level, can you conclude that
the mean length of all current long-distance calls is different from 12.44 minutes?
Solution Let /a be the mean length of all current long-distance calls placed through this
company and x be the corresponding mean for the sample. From the given information,
n = 150, x = 13.71 minutes, and cr = 2.65 minutes
We are to test whether or not the mean length of all current long-distance calls is different
from 12.44 minutes. The significance level a is .02; that is, the probability of rejecting the
null hypothesis when it actually is true should not exceed .02. This is the probability of mak-
ing a Type I error. We perform the test of hypothesis using the five steps as follows.
Conducting a two-tailed
test of hypothesis about fx: a
known and n > 30.
396 Chapter 9 Hypothesis Tests About the Mean and Proportion
Step 1. State the null and alternative hypotheses.
Notice that we are testing to find whether or not the mean length of all current long-
distance calls is different from 12.44 minutes. We write the null and alternative hypotheses
as follows.
H : ijl = 12.44 (The mean length of all current long-distance calls is 12.44 minutes.)
H x : ijl ¥= 12.44 (The mean length of all current long-distance calls is different from
12.44 minutes.)
Step 2. Select the distribution to use.
Here, the population standard deviation a is known, and the sample size is large (n > 30).
Hence, the sampling distribution of x is (approximately) normal (due to the Central Limit
Theorem) with its mean equal to /x and the standard deviation equal to o> = cr/Vn. Conse-
quently, we will use the normal distribution to perform the test of this example.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .02. The sign in the alternative hypothesis indicates that the test
is two-tailed with two rejection regions, one in each tail of the normal distribution curve of x.
Because the total area of both rejection regions is .02 (the significance level), the area of the
rejection region in each tail is .01; that is,
Area in each tail = a/2 = .02/2 = .01
These areas are shown in Figure 9.9. Two critical points in this figure separate the two rejec-
tion regions from the nonrejection region. Next, we find the z values for the two critical points
using the area of the rejection region. To find the z values for these critical points, we look
for .0100 and .9900 areas in the normal distribution table. From Table IV, the z values of the
two critical points, as shown in Figure 9.9, are approximately —2.33 and 2.33.
a/2 = .
51 /
a/
2 = .01
Reject h
\x = 12.44
-< Do not reject H >-
X
Reject H Q
-2.33 2.33
1 Two critical values of; 1
Figure 9.9
Step 4. Calculate the value of the test statistic.
The decision to reject or not to reject the null hypothesis will depend on whether the ev-
idence from the sample falls in the rejection or the nonrejection region. If the value of x falls
in either of the two rejection regions, we reject H Q . Otherwise, we do not reject H . The value
of x obtained from the sample is called the observed value of x. To locate the position of
x = 13.71 on the sampling distribution curve of x in Figure 9.9, we first calculate the z value
for x = 13.71. This is called the value of the test statistic. Then, we compare the value of
the test statistic with the two critical values of z, —2.33 and 2.33, shown in Figure 9.9. If the
value of the test statistic is between —2.33 and 2.33, we do not reject H . If the value of the
test statistic is either greater than 2.33 or less than —2.33, we reject H .
9.2 Hypothesis Tests About fj.: a Known 397
Calculating the Value of the Test Statistic When using the normal distribution, the value of the
test statistic z for x for a test of hypothesis about jx is computed as follows:
x — fl
where = — 7=
Vn
This value of z for x is also called the observed value of z.
The value of x from the sample is 13.71. We calculate the z value as follows:
a 2.65
From H
.21637159
5.87
The value of /j, in the calculation of the z value is substituted from the null hypothesis. The
value of z = 5.87 calculated for x is called the computed value of the test statistic z. This is
the value of z that corresponds to the value of x observed from the sample. It is also called
the observed value of z.
Step 5. Make a decision.
In the final step we make a decision based on the location of the value of the test statistic
z computed for x in Step 4. This value of z = 5.87 is greater than the critical value of z = 2.33,
and it falls in the rejection region in the right tail in Figure 9.9. Hence, we reject H and con-
clude that based on the sample information, it appears that the mean length of all such calls
is not equal to 12.44 minutes.
By rejecting the null hypothesis, we are stating that the difference between the sample mean,
x = 13.71 minutes, and the hypothesized value of the population mean, /jl = 12 .44 minutes, is
too large and may not have occurred because of chance or sampling error alone. This difference
seems to be real and, hence, the mean length of all such calls is different from 12.44 minutes.
Note that the rejection of the null hypothesis does not necessarily indicate that the mean length
of all such calls is definitely different from 12.44 minutes. It simply indicates that there is strong
evidence (from the sample) that the mean length of such calls is not equal to 12.44 minutes.
There is a possibility that the mean length of all such calls is equal to 12.44 minutes, but by the
luck of the draw we selected a sample with a mean that is too far from the hypothesized mean
of 12.44 minutes. If so, we have wrongfully rejected the null hypothesis H . This is a Type I
error and its probability is .02 in this example. I
We can use the />value approach to perform the test of hypothesis in Example 9-3. In this
example, the test is two-tailed. The />value is equal to twice the area under the sampling distri-
bution of x to the right of x = 13.71. As calculated in Step 4 above, the z value for x = 13.71 is
5.87. From the normal distribution table, the area to the right of z = 5.87 is (approximately) zero.
Hence, the p-value is zero. (If you use technology, you will obtain the p-value of .0000.) As we
know from earlier discussions, we will reject the null hypothesis for any a (significance level) that
is greater than the p-value. Consequently, in this example, we will reject the null hypothesis for
any a > 0. Since a = .02 here, which is greater than zero, we reject the null hypothesis.
■ EXAMPLE 9-4
The mayor of a large city claims that the average net worth of families living in this city is at
least $300,000. A random sample of 25 families selected from this city produced a mean net
worth of $288,000. Assume that the net worths of all families in this city have a normal dis-
tribution with the population standard deviation of $80,000. Using the 2.5% significance level,
can you conclude that the mayor's claim is false?
Conducting a left-tailed test
of hypothesis about fj. : a known,
n < 30, and population normal.
398 Chapter 9 Hypothesis Tests About the Mean and Proportion
Solution Let fi be the mean net worth of families living in this city and x be the correspon-
ding mean for the sample. From the given information,
n = 25, x = $288,000, and a = $80,000
The significance level is a = .025.
Step 1. State the null and alternative hypotheses.
We are to test whether or not the mayor's claim is false. The mayor's claim is that the av-
erage net worth of families living in this city is at least $300,000. Hence, the null and alter-
native hypotheses are as follows:
H : fjb > $300,000 (The mayor's claim is true. The mean net worth
is at least $300,000.)
H x : j± < $300,000 (The mayor's claim is false. The mean net worth
is less than $300,000.)
Step 2. Select the distribution to use.
Here, the population standard deviation cr is known, the sample size is small (n < 30), but
the population distribution is normal. Hence, the sampling distribution of x is normal with its
mean equal to jx and the standard deviation equal to cr^ = cr/Vn. Consequently, we will use
the normal distribution to perform the test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .025. The < sign in the alternative hypothesis indicates that the
test is left-tailed with the rejection region in the left tail of the sampling distribution curve of
x. The critical value of z, obtained from the normal table for .0250 area in the left tail, is
— 1.96, as shown in Figure 9.10.
a = .025-i
Reject H Q
jj= $300,000 x
-< Do not reject H
-1 .96
t
' — Critical value of z
Figure 9.10
Step 4. Calculate the value of the test statistic.
The value of the test statistic z for x = $288,000 is calculated as follows:
a 80,000
$16,000
'25
4T
- From H
_ x - (ju _ 288,000 - 300,000 _ _
Z ~ <t x 16,000
Step 5. Make a decision.
The value of the test statistic z = —.75 is greater than the critical value of z = —1.96, and
it falls in the nonrejection region. As a result, we fail to reject H . Therefore, we can state that
based on the sample information, it appears that the mean net worth of families in this city is
not less than $300,000. Note that we are not concluding that the mean net worth is definitely
not less than $300,000. By not rejecting the null hypothesis, we are saying that the information
USA TODAY Snapshots®
How crashes affect auto premiums
Average annual auto insurance premiums rise
with each at-fauk traffic accident:
$137
| $1,689
$2,041
By Aunt Carey and Keith Simmons, USA TODAY
The above chart shows the average cost of annual auto insurance premiums in 2008 for drivers with no
accidents, one accident, two accidents, and so on. According to the information in this chart, the average
annual auto insurance premium for drivers with no accidents was $1387 in 2008. It was $1689 for drivers
with one accident. Suppose these average premiums are true for all drivers in each category for the year
2008. Then we can check if the current average premium for any single category listed in the chart is higher
than the one in 2008. For example, suppose we want to check if the current average auto insurance pre-
mium for drivers with no accidents is higher than $1387. Suppose we take a random sample of 1600 driv-
ers with no accidents and find out that their current average annual auto insurance premium is $1450. As-
sume that the standard deviation of auto insurance premiums for all current drivers with no accidents is
$90, and the significance level is 1%. The test is right-tailed. The null and alternative hypotheses are
Hg-./j. = $1387
H,:^>$1387
Here, n = 1600, x = $1450, a = $90, and a = .01. The population standard deviation is known and the
sample is large. Hence, we can use the normal distribution to perform this test. Using the normal distribu-
tion to perform the test, the critical values of z for .01 area in the right tail of the normal curve is 2.33. We
find the observed value of z as follows.
HOW
CRASHES
AFFECT
AUTO
PREMIUMS
(T
90
Vn VT600
X - fJL
$2.25
1450 - 1387
2^25
28.00
The value of the test statistic z = 28.00 for x is larger than the critical value of z = 2.33 and it falls in
the rejection region. Consequently, we reject H and conclude that the current average auto insurance pre-
mium for drivers with no accidents is higher than $1387.
To use the p-value approach, we find the area under the normal curve to the right of z = 28.00 from
the normal distribution table. This area is zero. Therefore, the p-value (rounded to four decimal places) is
.0000. Since a = .01 is larger than .0000, we reject the null hypothesis.
Source: The chart reproduced with
permission from USA TODAY, August 1 2,
2009. Copyright © 2009, USA TODAY.
obtained from the sample is not strong enough to reject the null hypothesis and to conclude
that the mayor's claim is false. H
We can use the /7-value approach to perform the test of hypothesis in Example 9-4. In this
example, the test is left-tailed. The p-value is given by the area under the sampling distribution of
x to the left of x = $288,000. As calculated in Step 4 above, the z value for x = $288,000 is -.75.
399
400 Chapter 9 Hypothesis Tests About the Mean and Proportion
From the normal distribution table, the area to the left of z = — .75 is .2266. Hence, the p- value
is .2266. We will reject the null hypothesis for any a (significance level) that is greater than the
p-value. Consequently, we will reject the null hypothesis in this example for any a > .2266. Since
in this example a = .025, which is less than .2266, we fail to reject the null hypothesis.
In studies published in various journals, authors usually use the terms significantly differ-
ent and not significantly different when deriving conclusions based on hypothesis tests. These
terms are short versions of the terms statistically significantly different and statistically not sig-
nificantly different. The expression significantly different means that the difference between the
observed value of the sample mean x and the hypothesized value of the population mean p is
so large that it probably did not occur because of the sampling error alone. Consequently, the
null hypothesis is rejected. In other words, the difference between x and p is statistically sig-
nificant. Thus, the statement significantly different is equivalent to saying that the null hypoth-
esis is rejected. In Example 9-3, we can state as a conclusion that the observed value of x = 13.71
minutes is significantly different from the hypothesized value of p = 12.44 minutes. That is,
the mean length of all current long-distance calls is different from 12.44 minutes.
On the other hand, the statement not significantly different means that the difference between
the observed value of the sample mean x and the hypothesized value of the population mean p
is so small that it may have occurred just because of chance. Consequently, the null hypothesis
is not rejected. Thus, the expression not significantly different is equivalent to saying that we fail
to reject the null hypothesis. In Example 9-4, we can state as a conclusion that the observed
value of x = $288,000 is not significantly less than the hypothesized value of p = $300,000. In
other words, the current mean net worth of households in this city is not less than $300,000.
EXERCISES
CONCEPTS AND PROCEDURES
9.11 What are the five steps of a test of hypothesis using the critical value approach? Explain briefly.
9.12 What does the level of significance represent in a test of hypothesis? Explain.
9.13 By rejecting the null hypothesis in a test of hypothesis example, are you stating that the alternative
hypothesis is true?
9.14 What is the difference between the critical value of z and the observed value of z?
9.15 Briefly explain the procedure used to calculate the p-value for a two-tailed and for a one-tailed test,
respectively.
9.16 Find the p-value for each of the following hypothesis tests.
a. H : p = 23, H x : p + 23, n = 50, x = 21.25, a = 5
b. H : p = 15, Hi. p < 15, n = 80, 5=13.25, tr = 5.5
c. H : p = 38, Hi. p > 38, n = 35, x = 40.25, a = 7.2
9.17 Find the p-value for each of the following hypothesis tests.
a. H : p = 46, H x : p + 46, n = 40, x = 49.60, a = 9.7
b. H :p = 26, Hi. p < 26, n = 33, 5 = 24.30, a = 4.3
c. H : p = 18, Hi. p > 18, n = 55, x = 20.50, a = 7.8
9.18 Consider H : p = 29 versus H y : p + 29. A random sample of 25 observations taken from this pop-
ulation produced a sample mean of 25.3. The population is normally distributed with cr = 8.
a. Calculate the p-value.
b. Considering the p-value of part a, would you reject the null hypothesis if the test were made at
the significance level of .05?
c. Considering the p-value of part a, would you reject the null hypothesis if the test were made at
the significance level of .01?
9.19 Consider H : p = 72 versus H{. p > 72. A random sample of 16 observations taken from this pop-
ulation produced a sample mean of 75.2. The population is normally distributed with cr = 6.
a. Calculate the p-value.
b. Considering the p-value of part a, would you reject the null hypothesis if the test were made at
the significance level of .01?
c. Considering the p-value of part a, would you reject the null hypothesis if the test were made at
the significance level of .025?
9.2 Hypothesis Tests About /jl: a Known 401
9.20 For each of the following examples of tests of hypotheses about fi, show the rejection and nonrejec-
tion regions on the sampling distribution of the sample mean assuming that it is normal.
a. A two-tailed test with a = .05 and n = 40
b. A left-tailed test with a = .01 and n = 20
c. A right-tailed test with a = .02 and n = 55
9.21 For each of the following examples of tests of hypotheses about /jl, show the rejection and nonrejec-
tion regions on the sampling distribution of the sample mean assuming it is normal.
a. A two-tailed test with a = .01 and n = 100
b. A left-tailed test with a = .005 and n = 27
c. A right-tailed test with a = .025 and n = 36
9.22 Consider the following null and alternative hypotheses:
H : /jl = 25 versus H{. ii + 25
Suppose you perform this test at a = .05 and reject the null hypothesis. Would you state that the dif-
ference between the hypothesized value of the population mean and the observed value of the sample
mean is "statistically significant" or would you state that this difference is "statistically not significant"?
Explain.
9.23 Consider the following null and alternative hypotheses:
H Q : /jl = 60 versus H{. /jl > 60
Suppose you perform this test at a = .01 and fail to reject the null hypothesis. Would you state that the
difference between the hypothesized value of the population mean and the observed value of the sample
mean is "statistically significant" or would you state that this difference is "statistically not significant"?
Explain.
9.24 For each of the following significance levels, what is the probability of making a Type I error?
a. a = .025 b. a = .05 c. a = .01
9.25 For each of the following significance levels, what is the probability of making a Type I error?
a. a = .10 b. a = .02 c. a = .005
9.26 A random sample of 80 observations produced a sample mean of 86.50. Find the critical and ob-
served values of z for each of the following tests of hypothesis using a = .10. The population standard
deviation is known to be 7.20.
a. H : /jl = 91 versus H x ; /x + 91
b. H a : yii = 91 versus H { : ^ < 91
9.27 A random sample of 18 observations produced a sample mean of 9.24. Find the critical and observed
values of z for each of the following tests of hypothesis using a = .05. The population standard deviation
is known to be 5.40 and the population distribution is normal.
a. H : fx = 8.5 versus H^, fi + 8.5
b. H Q : yii = 8.5 versus H t : /x > 8.5
9.28 Consider the null hypothesis H : fi = 625. Suppose that a random sample of 29 observations is taken
from a normally distributed population with a = 32. Using a significance level of .01, show the rejection
and nonrejection regions on the sampling distribution curve of the sample mean and find the critical value(s)
of z when the alternative hypothesis is as follows.
a. Hi. fx + 625 b. H x : /j, > 625 c. H{. ^ < 625
9.29 Consider the null hypothesis H : fi = 5. A random sample of 140 observations is taken from a pop-
ulation with <j = 17. Using a = .05, show the rejection and nonrejection regions on the sampling distri-
bution curve of the sample mean and find the critical value(s) of z for the following.
a. a right-tailed test b. a left-tailed test c. a two-tailed test
9.30 Consider H : jx = 100 versus ffj: ti =h 100.
a. A random sample of 64 observations produced a sample mean of 98. Using a = .01, would you
reject the null hypothesis? The population standard deviation is known to be 12.
b. Another random sample of 64 observations taken from the same population produced a sample
mean of 104. Using a = .01, would you reject the null hypothesis? The population standard
deviation is known to be 12.
Comment on the results of parts a and b.
402 Chapter 9 Hypothesis Tests About the Mean and Proportion
9.31 Consider H a : p = 45 versus H{. \x, < 45.
a. A random sample of 25 observations produced a sample mean of 41.8. Using a = .025, would you
reject the null hypothesis? The population is known to be normally distributed with cr = 6.
b. Another random sample of 25 observations taken from the same population produced a sample
mean of 43.8. Using a = .025, would you reject the null hypothesis? The population is known
to be normally distributed with a = 6.
Comment on the results of parts a and b.
9.32 Make the following tests of hypotheses.
a. H : fjL = 25, H t : /a * 25, n = 81, x = 28.5, a = 3, a = .01
b. H : fji = 12, Hi. fji. < 12, n = 45, x = 11.25, a = 4.5, a = .05
c. H : p = 40, Hi. fi > 40, n = 100, x = 47, cr = 7, a = .10
9.33 Make the following tests of hypotheses.
a. H : p = 80, Hi. fx * 80, n = 33, x = 76.5, a = 15, a = .10
b. # : n = 32, H t : /a < 32, n = 75, 3c = 26.5, o- = 7.4, a = .01
c. H : p = 55, Hi. p > 55, n = 40, 3c = 60.5, cr = 4, a = .05
■ APPLICATIONS
9.34 A consumer advocacy group suspects that a local supermarket's 10-ounce packages of cheddar cheese
actually weigh less than 10 ounces. The group took a random sample of 20 such packages and found that
the mean weight for the sample was 9.955 ounces. The population follows a normal distribution with the
population standard deviation of .15 ounce.
a. Find the p-value for the test of hypothesis with the alternative hypothesis that the mean weight
of all such packages is less than 10 ounces. Will you reject the null hypothesis at a = .01?
b. Test the hypothesis of part a using the critical- value approach and a = .01.
9.35 The manufacturer of a certain brand of auto batteries claims that the mean life of these batteries is
45 months. A consumer protection agency that wants to check this claim took a random sample of 24 such
batteries and found that the mean life for this sample is 43.05 months. The lives of all such batteries have
a normal distribution with the population standard deviation of 4.5 months.
a. Find the p-value for the test of hypothesis with the alternative hypothesis that the mean life of
these batteries is less than 45 months. Will you reject the null hypothesis at a = .025?
b. Test the hypothesis of part a using the critical- value approach and a = .025.
9.36 A study claims that all adults spend an average of 14 hours or more on chores during a weekend. A
researcher wanted to check if this claim is true. A random sample of 200 adults taken by this researcher
showed that these adults spend an average of 14.65 hours on chores during a weekend. The population
standard deviation is known to be 3.0 hours.
a. Find the p-value for the hypothesis test with the alternative hypothesis that all adults spend more
than 14 hours on chores during a weekend. Will you reject the null hypothesis at a = .01?
b. Test the hypothesis of part a using the critical- value approach and a = .01.
9.37 A May 8, 2008, report on National Public Radio (www.npr.org) noted that the average age of first-
time mothers in the United States is slightly higher than 25 years. Suppose that a recently taken random
sample of 57 first-time mothers from Missouri produced an average age of 23.90 years and that the pop-
ulation standard deviation is known to be 4.80 years.
a. Find the p-value for the test of hypothesis with the alternative hypothesis that the current mean
age of all first-time mothers in Missouri is less than 25 years. Will you reject the null hypothesis
at a = .025?
b. Test the hypothesis of part a using the critical- value approach and a = .025.
9.38 The Bath Heritage Days, which take place in Bath, Maine, have been popular for, among other things,
an eating contest. In 2009, the contest switched from blueberry pie to a Whoopie Pie
(www.timesrecord.com), which consists of two large, chocolate cake-like cookies filled with a large amount
of vanilla cream. Sixty-five randomly selected adults are chosen to eat a 1 -pound Whoopie Pie, and the
average time for 59 adults (out of these 65) is 127.10 seconds. Based on other Whoopie Pie-eating con-
tests throughout the United States, suppose that the standard deviation of the times taken by all adults to
consume 1-pound Whoopie pies are known to be 23.80 seconds.
a. Find the p-value for the test of hypothesis with the alternative hypothesis that the mean time to
eat a 1-pound Whoopie Pie is more than 2 minutes. Will you reject the null hypothesis at a = .01?
Explain. What if a = .02?
b. Test the hypothesis of part a using the critical-value approach. Will you reject the null hypothe-
sis at a = .01? What if a = .02?
9.2 Hypothesis Tests About ft: a Known
403
9.39 A telephone company claims that the mean duration of all long-distance phone calls made by its res-
idential customers is 10 minutes. A random sample of 100 long-distance calls made by its residential cus-
tomers taken from the records of this company showed that the mean duration of calls for this sample is
9.20 minutes. The population standard deviation is known to be 3.80 minutes.
a. Find the p-value for the test that the mean duration of all long-distance calls made by residential
customers is different from 10 minutes. If a = .02, based on this p-vafue, would you reject the
null hypothesis? Explain. What if a = .05?
b. Test the hypothesis of part a using the critical- value approach and a = .02. Does your conclusion
change if a = .05?
9.40 Lazurus Steel Corporation produces iron rods that are supposed to be 36 inches long. The ma-
chine that makes these rods does not produce each rod exactly 36 inches long. The lengths of the rods
are normally distributed, and they vary slightly. It is known that when the machine is working prop-
erly, the mean length of the rods is 36 inches. The standard deviation of the lengths of all rods pro-
duced on this machine is always equal to .035 inch. The quality control department at the company
takes a sample of 20 such rods every week, calculates the mean length of these rods, and tests the null
hypothesis, /x = 36 inches, against the alternative hypothesis, /j, + 36 inches. If the null hypothesis is
rejected, the machine is stopped and adjusted. A recent sample of 20 rods produced a mean length of
36.015 inches.
a. Calculate the p-value for this test of hypothesis. Based on this p-value, will the quality control
inspector decide to stop the machine and adjust it if he chooses the maximum probability of a
Type I error to be .02? What if the maximum probability of a Type I error is .10?
b. Test the hypothesis of part a using the critical-value approach and a = .02. Does the machine
need to be adjusted? What if a = .10?
9.41 At Farmer's Dairy, a machine is set to fill 32-ounce milk cartons. However, this machine does not
put exactly 32 ounces of milk into each carton; the amount varies slightly from carton to carton but has
a normal distribution. It is known that when the machine is working properly, the mean net weight of
these cartons is 32 ounces. The standard deviation of the milk in all such cartons is always equal to .15
ounce. The quality control inspector at this company takes a sample of 25 such cartons every week, cal-
culates the mean net weight of these cartons, and tests the null hypothesis, /x = 32 ounces, against the
alternative hypothesis, /j, + 32 ounces. If the null hypothesis is rejected, the machine is stopped and ad-
justed. A recent sample of 25 such cartons produced a mean net weight of 31.93 ounces.
a. Calculate the p-value for this test of hypothesis. Based on this p-value, will the quality control
inspector decide to stop the machine and readjust it if she chooses the maximum probability of a
Type I error to be .01? What if the maximum probability of a Type I error is .05?
b. Test the hypothesis of part a using the critical-value approach and a = .01. Does the machine
need to be adjusted? What if a = .05?
9.42 The Environmental Protection Agency (EPA) recommends that the sodium content in public
water supplies should be no more than 20 mg per liter (http://www.disabledworld.com/artman/publish/
sodiumwatersupply.shtml). Forty samples were taken from a large reservoir, and the amount of sodium in
each sample was measured. The sample average was 23.5 mg per liter. Assume that the population stan-
dard deviation is 5.6 mg per liter. The EPA is interested in knowing whether the average sodium content
for the entire reservoir exceeds the recommended level. If so, the communities served by the reservoir will
have to be made aware of the violation.
a. Find the p-value for the test of hypothesis. Based on this p-value, would the communities need
to be informed of an excessive average sodium level if the maximum probability of a Type I
error is to be .05? What if the maximum probability of a Type I error is to be .01?
b. Test the hypothesis of part a using the critical- value approach and a = .05. Would the commu-
nities need to be notified? What if a = .01? What if a is zero?
9.43 Records in a three-county area show that in the last few years, Girl Scouts sell an average of 47.93
boxes of cookies per year, with a population standard deviation of 8.45 boxes per year. Fifty randomly se-
lected Girl Scouts from the region sold an average of 46.54 boxes this year. Scout leaders are concerned
that the demand for Girl Scout cookies may have decreased.
a. Test at the 10% significance level whether the average number of boxes of cookies sold by all
Girl Scouts in the three-county area is lower than the historical average.
b. What will your decision be in part a if the probability of a Type I error is zero? Explain.
9.44 A journalist claims that all adults in her city spend an average of 30 hours or more per month on
general reading, such as newspapers, magazines, novels, and so forth. A recent sample of 25 adults from
this city showed that they spend an average of 27 hours per month on general reading. The population
of such times is known to be normally distributed with the population standard deviation of 7 hours.
404
Chapter 9 Hypothesis Tests About the Mean and Proportion
a. Using the 2.5% significance level, would you conclude that the mean time spent per month on such
reading by all adults in this city is less than 30 hours? Use both procedures — the p- value approach
and the critical value approach.
b. Make the test of part a using the 1% significance level. Is your decision different from that of
part a? Comment on the results of parts a and b.
9.45 A study claims that all homeowners in a town spend an average of 8 hours or more on house clean-
ing and gardening during a weekend. A researcher wanted to check if this claim is true. A random sam-
ple of 20 homeowners taken by this researcher showed that they spend an average of 7.68 hours on such
chores during a weekend. The population of such times for all homeowners in this town is normally dis-
tributed with the population standard deviation of 2.1 hours.
a. Using the 1% significance level, can you conclude that the claim that all homeowners spend an
average of 8 hours or more on such chores during a weekend is false? Use both approaches.
b. Make the test of part a using a 2.5% significance level. Is your decision different from the one in
part a? Comment on the results of parts a and b.
9.46 A company claims that the mean net weight of the contents of its All Taste cereal boxes is at least
18 ounces. Suppose you want to test whether or not the claim of the company is true. Explain briefly how
you would conduct this test using a large sample. Assume that a = .25 ounce.
9.3 Hypothesis Tests About \x\ a Not Known
This section explains how to perform a test of hypothesis about the population mean jx when the
population standard deviation a is not known. Here, again, there are three possible cases as follows.
Case I. If the following three conditions are fulfilled:
1. The population standard deviation <x is not known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is normally distributed,
then we use the t distribution to perform a test of hypothesis about /jl.
Case II. If the following two conditions are fulfilled:
1. The population standard deviation cr is not known
2. The sample size is large (i.e., n > 30),
then again we use the t distribution to perform a test of hypothesis about fi.
Case III. If the following three conditions are fulfilled:
1. The population standard deviation a is not known
2. The sample size is small (i.e., n < 30)
3. The population from which the sample is selected is not normally distributed (or the
shape of its distribution is unknown),
then we use a nonparametric method to perform a test of hypothesis about /jl.
The following chart summarizes the above three cases.
CT Is Not Known
1
1
1
Case I
1. n < 30
2. Population is normal
Case II
n 2: 30
Case HI
1. n < 30
2. Population is not normal
Use the t distribution to
perform a test of hypothesis
about fjL
Use a nonparametric method
to perform a test of hypothesis
about ju,
9.3 Hypothesis Tests About /u,: a Not Known 405
Below we discuss Cases I and II and learn how to use the t distribution to perform a test
of hypothesis about jx when cr is not known. When the conditions mentioned for Case I or Case
II are satisfied, the random variable
x - (X s
t = where St. = — t=
s- x Vn
has a t distribution. Here, the t is called the test statistic to perform a test of hypothesis about
a population mean jx.
Test Statistic The value of the test statistic t for the sample mean x is computed as
x — fX s
t = where Sy = — 7=
s x * Vn
The value of t calculated for x by using this formula is also called the observed value of t.
In Section 9.2, we discussed two procedures, the />value approach and the critical-value
approach, to test hypotheses about /X when cr is known. In this section also we will use these
two procedures to test hypotheses about jx when cr is not known. The steps used in these pro-
cedures are the same as in Section 9.2. The only difference is that we will be using the t distri-
bution in place of the normal distribution.
1 . The p-Value Approach
To use the /7-value approach to perform a test of hypothesis about jx using the t distribution,
we will use the same four steps that we used in such a procedure in Section 9.2. Although the
p-value can be obtained by using any technology very easily, we can use Table V of Appen-
dix C to find a range for the /7-value when technology is not available. Note that when using
the t distribution and Table V, we cannot find the exact /7-value but only a range within which
it falls.
Examples 9-5 and 9-6 illustrate the /7-value procedure to test a hypothesis about jx using
the t distribution.
■ EXAMPLE 9-5
A psychologist claims that the mean age at which children start walking is 12.5 months. Carol
wanted to check if this claim is true. She took a random sample of 18 children and found that
the mean age at which these children started walking was 12.9 months with a standard
deviation of .80 month. It is known that the ages at which all children start walking are ap-
proximately normally distributed. Find the /7-value for the test that the mean age at which all
children start walking is different from 12.5 months. What will your conclusion be if the sig-
nificance level is 1%?
Solution Let fx be the mean age at which all children start walking, and let x be the corre-
sponding mean for the sample. From the given information,
n = 18, x = 12.9 months, and s = .80 month
The claim of the psychologist is that the mean age at which children start walking is 12.5 months.
To calculate the /7-value and to make the decision, we apply the following four steps.
Step 1. State the null and alternative hypotheses.
We are to test if the mean age at which all children start walking is different from 12.5 months.
Hence, the null and alternative hypotheses are
H : ix = 12.5 (The mean walking age is 12.5 months.)
H x : jx # 12.5 (The mean walking age is different from 12.5 months.)
Finding a p-value and making
a decision for a two-tailed test
of hypothesis about fi: o~ not
known, n < 30, and
population normal.
st
406 Chapter 9 Hypothesis Tests About the Mean and Proportion
Step 2. Select the distribution to use.
In this example, we do not know the population standard deviation cr, the sample size is
small (n < 30), and the population is normally distributed. Hence, it is Case I mentioned in
the beginning of this section. Consequently, we will use the t distribution to find the p-value
for this test.
Step 3. Calculate the p-value.
The # sign in the alternative hypothesis indicates that the test is two-tailed. To find the
p-value, first we find the degrees of freedom and the t value for x = 12.9 months. Then, the
p-value is equal to twice the area in the tail of the t distribution curve to the right of this t
value for x = 12.9 months. This p-value is shown in Figure 9.11. We find this p-value as
follows:
s .80
% = —j= = = .18856181
Vn Vl8
X — fJb
and df = n
Now we can find the range for the p-value. To do so, we go to Table V of Appendix C (the
t distribution table) and find the row of df = 17. In this row, we find the two values of t
that cover t = 2.121. From Table V, for df = 17, these two values of t are 2.110 and 2.567.
The test statistic t = 2.121 falls between these two values. Now look in the top row of this
table to find the areas in the tail of the t distribution curve that correspond to 2.110 and
2.567. These two areas are .025 and .01, respectively. In other words, the area in the up-
per tail of the t distribution curve for df = 17 and t = 2.110 is .025, and the area in the up-
per tail of the t distribution curve for df = 17 and t = 2.567 is .01. Because it is a two-
tailed test, the p-value for t = 2.121 is between 2(.025) = .05 and 2(.01) = .02, which can
be written as
.02 < p-value < .05
Note that by using Table V of Appendix C, we cannot find the exact p-value but only a range
for it. If we have access to technology, we can find the exact p-value by using technology. If
we use technology for this example, we will obtain a p-value of .049.
j- From H
12.9 - 12.5
= 2.121
.18856181
1 = 18 - 1 = 17
Step 4. Make a decision.
Thus, we can state that for any a greater than .05 (the upper limit of the p-value range),
we will reject the null hypothesis. For any a less than .02 (the lower limit of the p-value range),
we will not reject the null hypothesis. However, if a is between .02 and .05, we cannot make
a decision. Note that if we use technology, then the p-value we will obtain for this example is
.049, and we can make a decision for any value of a. For our example, a = .01, which is less
than the lower limit of the p-value range of .02. As a result, we fail to reject H and conclude
that the mean age at which all children start walking is not different from 12.5 months. As a
result, we can state that the difference between the hypothesized population mean and the sam-
ple mean is so small that it may have occurred because of sampling error. I
9.3 Hypothesis Tests About /u,: a Not Known 407
■ EXAMPLE 9-6
Grand Auto Corporation produces auto batteries. The company claims that its top-of-the-line
Never Die batteries are good, on average, for at least 65 months. A consumer protection agency
tested 45 such batteries to check this claim. It found that the mean life of these 45 batteries
is 63.4 months, and the standard deviation is 3 months. Find the p-value for the test that the
mean life of all such batteries is less than 65 months. What will your conclusion be if the sig-
nificance level is 2.5%?
Solution Let /jl be the mean life of all such auto batteries, and let x be the corresponding
mean for the sample. From the given information,
n = 45, x = 63.4 months, and 5 = 3 months
The claim of the company is that the mean life of these batteries is at least 65 months. To cal-
culate the p-value and to make the decision, we apply the following four steps.
Step 1. State the null and alternative hypotheses.
We are to test if the mean life of these batteries is at least 65 months. Hence, the null and
alternative hypotheses are
H : j± > 65 (The mean life of batteries is at least 65 months.)
H x : /a < 65 (The mean life of batteries is less than 65 months.)
Step 2. Select the distribution to use.
In this example, we do not know the population standard deviation cr, and the sample size
is large (n > 30). Hence, it is Case II mentioned in the beginning of this section. Conse-
quently, we will use the t distribution to find the p-value for this test.
Step 3. Calculate the p-value.
The < sign in the alternative hypothesis indicates that the test is left-tailed. To find the
p-value, first we find the degrees of freedom and the t value for x = 63.4 months. Then, the
p-value is given by the area in the tail of the t distribution curve to the left of this t value for
x = 63.4 months. This p-value is shown in Figure 9.12. We find this p-value as follows:
= .44721360
j- From H
x- fji 63.4 - 65
t = = = -3.578
s 7x .44721360
and d/=«-l=45-l=44
Now we can find the range for the p-value. To do so, we go to Table V of Appendix C (the
t distribution table) and find the row of df = 44. In this row, we find the two values of t that
cover t = 3.578. Note that we use the positive value of the test statistic t, although our test
statistic has a negative value. From Table V, for df = 44, the largest value of t is 3.286, for
which the area in the tail of the t distribution is .001. This means that the area to the left of
t = -3.286 is .001. Because -3.578 is smaller than -3.286, the area to the left of t = -3.578
Finding a p-value and
making a decision for a left-
tailed test of hypothesis about
fi: o~ not known and n > 30.
408 Chapter 9 Hypothesis Tests About the Mean and Proportion
Com
Conducting a two-tailed test
of hypothesis about fi:
a unknown, n < 30, and
population normal.
is smaller than .001. Therefore, the p-value for t = —3.578 is less than .001, which can be
written as
p-value < .001
Thus, here the p-value has only the upper limit of .001. In other words, the /7-value for this
example is less than .001. If we use technology for this example, we will obtain a p- value
of .000.
Step 4. Make a decision.
Thus, we can state that for any a greater than .001 (the upper limit of the /7-value range),
we will reject the null hypothesis. For our example a = .025, which is greater than the upper
limit of the ;?-value of .001. As a result, we reject H and conclude that the mean life of
such batteries is less than 65 months. Therefore, we can state that the difference between the
hypothesized population mean of 65 months and the sample mean of 63.4 is too large to be
attributed to sampling error alone. I
2. The Critical-Value Approach
As mentioned in Section 9.2, this procedure is also called the traditional or classical approach.
In this procedure, we have a predetermined value of the significance level a. The value of a
gives the total area of the rejection region(s). First we find the critical value(s) of t from the
t distribution table in Appendix C for the given degrees of freedom and the significance level.
Then we find the value of the test statistic t for the observed value of the sample statistic x.
Finally we compare these two values and make a decision. Remember, if the test is one-tailed,
there is only one critical value of f, and it is obtained by using the value of a, which gives
the area in the left or right tail of the t distribution curve, depending on whether the test is
left-tailed or right-tailed, respectively. However, if the test is two-tailed, there are two criti-
cal values of f, and they are obtained by using a/2 area in each tail of the t distribution curve.
The value of the test statistic t is obtained as mentioned earlier in this section.
Examples 9-7 and 9-8 describe the procedure to test a hypothesis about p using the critical-
value approach and the t distribution.
■ EXAMPLE 9-7
Refer to Example 9-5. A psychologist claims that the mean age at which children start walking
is 12.5 months. Carol wanted to check if this claim is true. She took a random sample of 18 chil-
dren and found that the mean age at which these children started walking was 12.9 months with
a standard deviation of .80 month. Using the 1% significance level, can you conclude that the
mean age at which all children start walking is different from 12.5 months? Assume that the
ages at which all children start walking have an approximately normal distribution.
Solution Let p be the mean age at which all children start walking, and let x be the corre-
sponding mean for the sample. Then, from the given information,
n = 18, x = 12.9 months, s = .80 month, and a = .01
Step 1. State the null and alternative hypotheses.
We are to test if the mean age at which all children start walking is different from
12.5 months. The null and alternative hypotheses are
H : p = 12.5 (The mean walking age is 12.5 months.)
H\. p ¥= 12.5 (The mean walking age is different from 12.5 months.)
Step 2. Select the distribution to use.
In this example, the population standard deviation cr is not known, the sample size is small
(n < 30), and the population is normally distributed. Hence, it is Case I mentioned in the
beginning of this section. Consequently, we will use the t distribution to perform the test in
this example.
9.3 Hypothesis Tests About /i: a Not Known 409
Step 3. Determine the rejection and nonrejection regions.
The significance level is .01. The # sign in the alternative hypothesis indicates that the test
is two-tailed and the rejection region lies in both tails. The area of the rejection region in each
tail of the t distribution curve is
Area in each tail = a/2 = .01/2 = .005
df = n - 1 = 18 - 1 = 17
From the t distribution table, the critical values of t for 17 degrees of freedom and .005 area in
each tail of the t distribution curve are —2.898 and 2.898. These values are shown in Figure 9.13.
Step 4. Calculate the value of the test statistic.
We calculate the value of the test statistic t for x = 12.9 as follows:
s .80
.18856181
<n V18
From H„
r
_/a _ 12.9 - 12.5
~~ .18856181
2.121
Step 5. Make a decision.
The value of the test statistic t = 2.121 falls between the two critical points, —2.898 and
2.898, which is the nonrejection region. Consequently, we fail to reject H . As a result, we can
state that the difference between the hypothesized population mean and the sample mean is
so small that it may have occurred because of sampling error. The mean age at which children
start walking is not different from 12.5 months. H
■ EXAMPLE 9-8
The management at Massachusetts Savings Bank is always concerned about the quality of
service provided to its customers. With the old computer system, a teller at this bank could
serve, on average, 22 customers per hour. The management noticed that with this service rate,
the waiting time for customers was too long. Recently the management of the bank installed
a new computer system, expecting that it would increase the service rate and consequently
make the customers happier by reducing the waiting time. To check if the new computer sys-
tem is more efficient than the old system, the management of the bank took a random sam-
ple of 70 hours and found that during these hours the mean number of customers served by
tellers was 27 per hour with a standard deviation of 2.5. Testing at the 1% significance level,
would you conclude that the new computer system is more efficient than the old computer
system?
Solution Let /x be the mean number of customers served per hour by a teller using the
new system, and let x be the corresponding mean for the sample. Then, from the given
information,
Conducting a right-tailed
test of hypothesis about fi: <j
unknown and n > 30.
n = 70 hours, x = 27 customers,
2.5 customers, and a
.01
410 Chapter 9 Hypothesis Tests About the Mean and Proportion
Step 1. State the null and alternative hypotheses.
We are to test whether or not the new computer system is more efficient than the old sys-
tem. The new computer system will be more efficient than the old system if the mean num-
ber of customers served per hour by using the new computer system is significantly more than
22; otherwise, it will not be more efficient. The null and alternative hypotheses are
H : )jl = 22 (The new computer system is not more efficient.)
H x \ [i > 22 (The new computer system is more efficient.)
Step 2. Select the distribution to use.
In this example, the population standard deviation a is not known and the sample size is
large (n > 30). Hence, it is Case II mentioned in the beginning of this section. Consequently,
we will use the t distribution to perform the test for this example.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .01. The > sign in the alternative hypothesis indicates that the test
is right-tailed and the rejection region lies in the right tail of the t distribution curve.
Area in the right tail = a = .01
df = n - 1 = 70 - 1 = 69
From the t distribution table, the critical value of t for 69 degrees of freedom and .01 area in
the right tail is 2.382. This value is shown in Figure 9.14.
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for I = 27 is calculated as follows:
s 2.5
s- x = ^ = = .29880715
Vb V70
I From H a
fi 27-22
16.733
s 7x .29880715
Step 5. Make a decision.
The value of the test statistic t = 16.733 is greater than the critical value of t = 2.382, and
it falls in the rejection region. Consequently, we reject H . As a result, we conclude that the
value of the sample mean is too large compared to the hypothesized value of the population
mean, and the difference between the two may not be attributed to chance alone. The mean
number of customers served per hour using the new computer system is more than 22. The new
computer system is more efficient than the old computer system. H
Note: What If the Sample Size Is Too Large?
In the above section when <x is not known, we used the t distribution to perform tests of hypothe-
sis about jx in Cases I and II. Note that in Case II, the sample size is large. If we have access to
technology, it does not matter how large (greater than 30) the sample size is, we can use the t dis-
tribution. However, if we are using the t distribution table (Table V of Appendix C), this may pose
9.3 Hypothesis Tests About /jl: a Not Known 41 1
a problem. Usually such a table only goes up to a certain number of degrees of freedom. For ex-
ample, Table V in Appendix C only goes up to 75 degrees of freedom. Thus, if the sample size is
larger than 76 (with df more than 75) here, we cannot use Table V to find the critical value(s) of t
to make a decision in this section. In such a situation when n is too large and is not included in the
t distribution table, there are two options:
1. Use the t value from the last row (the row of oo) in Table V of Appendix C.
2. Use the normal distribution as an approximation to the t distribution.
To use the normal distribution as an approximation to the t distribution to make a test of
hypothesis about /jl, the procedure is exactly like the one in Section 9.2, except that now we
will replace a by s, and o" Y by s„
Note that the t values obtained from the last row of the t distribution table are the same as
will be obtained from the normal distribution table for the same areas in the upper tail or lower
tail of the distribution. Again, note that here we can use the normal distribution as a conven-
ience and as an approximation, but if we can, we should use the t distribution by using tech-
nology. Exercises 9.70 and 9.71 at the end of this section present such situations.
EXERCISES
CONCEPTS AND PROCEDURES
9.47 Briefly explain the conditions that must hold true to use the t distribution to make a test of hypoth-
esis about the population mean.
9.48 For each of the following examples of tests of hypothesis about yii, show the rejection and nonrejec-
tion regions on the t distribution curve.
a. A two-tailed test with a = .02 and n = 20
b. A left- tailed test with a = .01 and n = 16
c. A right-tailed test with a = .05 and n = 18
9.49 For each of the following examples of tests of hypothesis about jjl, show the rejection and nonrejec-
tion regions on the t distribution curve.
a. A two- tailed test with a = .01 and n = 15
b. A left-tailed test with a = .005 and n = 25
c. A right-tailed test with a = .025 and n = 22
9.50 A random sample of 14 observations taken from a population that is normally distributed produced
a sample mean of 212.37 and a standard deviation of 16.35. Find the critical and observed values of t and
the ranges for the p- value for each of the following tests of hypotheses, using a = .10.
a. H : p = 205 versus H x : \x + 205
b. H : p = 205 versus H x : p > 205
9.51 A random sample of 8 observations taken from a population that is normally distributed produced a
sample mean of 44.98 and a standard deviation of 6.77. Find the critical and observed values of t and the
ranges for the p-value for each of the following tests of hypotheses, using a = .05.
a. H : \Xj — 50 versus H x : p + 50
b. H : p — 50 versus H t : /x < 50
9.52 Consider the null hypothesis H : /jl = 100. Suppose that a random sample of 35 observations is taken
from this population to perform this test. Using a significance level of .01, show the rejection and nonre-
jection regions and find the critical value(s) of t when the alternative hypothesis is as follows.
a. Hi. /jl + 100 b. H{. /jl > 100 c. H{. p, < 100
9.53 Consider the null hypothesis H : pL = 12.80. A random sample of 58 observations is taken from this
population to perform this test. Using a = .05, show the rejection and nonrejection regions on the sam-
pling distribution curve of the sample mean and find the critical value(s) of t for the following.
a. a right-tailed test b. a left-tailed test c. a two-tailed test
9.54 Consider H : /jl = 80 versus H^. fi 80 for a population that is normally distributed.
a. A random sample of 25 observations taken from this population produced a sample mean of 77
and a standard deviation of 8. Using a = .01, would you reject the null hypothesis?
b. Another random sample of 25 observations taken from the same population produced a sample
mean of 86 and a standard deviation of 6. Using a = .01, would you reject the null hypothesis?
Comment on the results of parts a and b.
412 Chapter 9 Hypothesis Tests About the Mean and Proportion
9.55 Consider H : p = 40 versus //,: /a > 40.
a. A random sample of 64 observations taken from this population produced a sample mean of 43
and a standard deviation of 5. Using a = .025, would you reject the null hypothesis?
b. Another random sample of 64 observations taken from the same population produced a sample
mean of 41 and a standard deviation of 7. Using a = .025, would you reject the null hypothesis?
Comment on the results of parts a and b.
9.56 Perform the following tests of hypothesis.
a. H : p, = 285, H x : p < 285, n = 55, x = 267.80, .v = 42.90, a = .05
b. H : p = 10.70, H x . p + 10.70, n = 47, x = 12.025, s = 4.90, a = .01
c. H : p = 147,500, H x : p < 147,500, n = 41, 3c = 141,812, .v = 22,972, a = .10
9.57 Perform the following tests of hypotheses for data coming from a normal distribution.
a. H : p = 94.80, H x ; p < 94.80, n =12, 3c = 92.87, s = 5.34, a = .10
b. H : p = 18.70, H x ; p + 18.70, n = 25, 3c = 20.05, s = 2.99, a = .05
c. « : p = 59, /a > 59, n = 7, 3f = 59.42, * = .418, a = .01
■ APPLICATIONS
9.58 The police that patrol a heavily traveled highway claim that the average driver exceeds the 65 miles
per hour speed limit by more than 10 miles per hour. Seventy-two randomly selected cars were clocked
by airplane radar. The average speed was 77.40 miles per hour, and the standard deviation of the speeds
was 5.90 miles per hour. Find the range for the p-value for this test. What will your conclusion be using
this p-value range and a = .02?
9.59 According to the Magazine Publishers of America (www.magazine.org), the average visit at the mag-
azines' Web sites during the fourth quarter of 2007 lasted 4.145 minutes. Forty-six randomly selected vis-
its to magazine's Web sites during the fourth quarter of 2009 produced a sample mean visit of 4.458 min-
utes, with a standard deviation of 1.14 minutes. Using the 10% significance level and the critical value
approach, can you conclude that the length of an average visit to these Web sites during the fourth quar-
ter of 2009 was longer than 4.145 minutes? Find the range for the p-value for this test. What will your
conclusion be using this p- value range and a = .10?
9.60 The president of a university claims that the mean time spent partying by all students at this uni-
versity is not more than 7 hours per week. A random sample of 40 students taken from this university
showed that they spent an average of 9.50 hours partying the previous week with a standard deviation of
2.3 hours. Test at the 2.5% significance level whether the president's claim is true. Explain your conclu-
sion in words.
9.61 The mean balance of all checking accounts at a bank on December 31, 2009, was $850. A random
sample of 55 checking accounts taken recently from this bank gave a mean balance of $780 with a stan-
dard deviation of $230. Using the 1% significance level, can you conclude that the mean balance of such
accounts has decreased during this period? Explain your conclusion in words. What if a = .025?
9.62 A soft-drink manufacturer claims that its 12-ounce cans do not contain, on average, more than 30 calo-
ries. A random sample of 64 cans of this soft drink, which were checked for calories, contained a mean
of 32 calories with a standard deviation of 3 calories. Does the sample information support the alternative
hypothesis that the manufacturer's claim is false? Use a significance level of 5%. Find the range for the
p-value for this test. What will your conclusion be using this p-value and a = .05?
9.63 In 2006, the average number of new single-family homes built per town in the state of Maine was
14.325 (www.mainehousing.org). Suppose that a random sample of 42 Maine towns taken in 2009 resulted
in an average of 13.833 new single-family homes built per town, with a standard deviation of 4.241 new
single-family homes. Using the 5% significance level, can you conclude that the average number of new
single-family homes per town built in 2009 in the state of Maine is significantly different from 14.325?
Use both the p- value and critical- value approaches.
9.64 A paint manufacturing company claims that the mean drying time for its paints is not longer than
45 minutes. A random sample of 20 gallons of paints selected from the production line of this company
showed that the mean drying time for this sample is 49.50 minutes with a standard deviation of 3 min-
utes. Assume that the drying times for these paints have a normal distribution.
a. Using the 1% significance level, would you conclude that the company's claim is true?
b. What is the Type I error in this exercise? Explain in words. What is the probability of making
such an error?
9.65 The manager of a restaurant in a large city claims that waiters working in all restaurants in his city earn
an average of $150 or more in tips per week. A random sample of 25 waiters selected from restaurants of
9.3 Hypothesis Tests About /*: <t Not Known 413
this city yielded a mean of $139 in tips per week with a standard deviation of $28. Assume that the weekly
tips for all waiters in this city have a normal distribution.
a. Using the 1 % significance level, can you conclude that the manager's claim is true? Use both approaches.
b. What is the Type I error in this exercise? Explain. What is the probability of making such an error?
9.66 A business school claims that students who complete a 3-month typing course can type, on average,
at least 1200 words an hour. A random sample of 25 students who completed this course typed, on aver-
age, 1 125 words an hour with a standard deviation of 85 words. Assume that the typing speeds for all stu-
dents who complete this course have an approximately normal distribution.
a. Suppose the probability of making a Type I error is selected to be zero. Can you conclude that the claim
of the business school is true? Answer without performing the five steps of a test of hypothesis.
b. Using the 5% significance level, can you conclude that the claim of the business school is true?
Use both approaches.
9.67 According to an estimate, 2 years ago the average age of all CEOs of medium-sized companies in
the United States was 58 years. Jennifer wants to check if this is still true. She took a random sample of
70 such CEOs and found their mean age to be 55 years with a standard deviation of 6 years.
a. Suppose that the probability of making a Type I error is selected to be zero. Can you conclude
that the current mean age of all CEOs of medium-sized companies in the Untied States is differ-
ent from 58 years?
b. Using the 1% significance level, can you conclude that the current mean age of all CEOs of
medium-sized companies in the United States is different from 58 years? Use both approaches.
9.68 A past study claimed that adults in America spent an average of 18 hours a week on leisure activi-
ties. A researcher wanted to test this claim. She took a sample of 12 adults and asked them about the time
they spend per week on leisure activities. Their responses (in hours) are as follows.
13.6 14.0 24.5 24.6 22.9 37.7 14.6 14.5 21.5 21.0 17.8 21.4
Assume that the times spent on leisure activities by all adults are normally distributed. Using the 10% sig-
nificance level, can you conclude that the average amount of time spent on leisure activities has changed?
(Hint: First calculate the sample mean and the sample standard deviation for these data using the formu-
las learned in Sections 3.1.1 and 3.2.2 of Chapter 3. Then make the test of hypothesis about fi.)
9.69 The past records of a supermarket show that its customers spend an average of $95 per visit at this
store. Recently the management of the store initiated a promotional campaign according to which each
customer receives points based on the total money spent at the store, and these points can be used to buy
products at the store. The management expects that as a result of this campaign, the customers should be
encouraged to spend more money at the store. To check whether this is true, the manager of the store took
a sample of 14 customers who visited the store. The following data give the money (in dollars) spent
by these customers at this supermarket during their visits.
109.15 136.01 107.02 116.15 101.53 109.29 110.79
94.83 100.91 97.94 104.30 83.54 67.59 120.44
Assume that the money spent by all customers at this supermarket has a normal distribution. Using the
5% significance level, can you conclude that the mean amount of money spent by all customers at this su-
permarket after the campaign was started is more than $95? (Hint: First calculate the sample mean and
the sample standard deviation for these data using the formulas learned in Sections 3.1.1 and 3.2.2 of
Chapter 3. Then make the test of hypothesis about /jl.)
9.70 According to an article in the Chicago Sun-Times (http://jaehakim.eom/ai1:icles/lifestyles/engage.htm),
the average length of an engagement that results in a marriage in the United States is 14 months. Suppose that
a random sample of 99 recently married Canadian couples had an average engagement length of 12.84 months,
with a sample standard deviation of 4.52 months. Does the sample information support the alternative hypoth-
esis that the average engagement length in Canada is different from 14 months, the average length in the United
States? Use a 10% significance level. Use both the p-value approach and the critical-value approach.
9.71 According to the University of Wisconsin Dairy Marketing and Risk Management Program (http://
future.aae.wisc.edu/index.html), the average retail price of 1 gallon of whole milk in the United States for
April 2009 was $3,084. A recent random sample of 80 retailers in the United States produced an average
milk price of $3,022 per gallon, with a standard deviation of $.274. Do the data provide significant evi-
dence at the 1% level to conclude that the current average price of 1 gallon of milk in the United States
is lower than the April 2009 average of $3,084?
*9.72 The manager of a service station claims that the mean amount spent on gas by its customers is
$15.90 per visit. You want to test if the mean amount spent on gas at this station is different from $15.90
per visit. Briefly explain how you would conduct this test when a is not known.
414 Chapter 9 Hypothesis Tests About the Mean and Proportion
*9.73 A tool manufacturing company claims that its top-of-the-line machine that is used to manufacture
bolts produces an average of 88 or more bolts per hour. A company that is interested in buying this ma-
chine wants to check this claim. Suppose you are asked to conduct this test. Briefly explain how you would
do so when cr is not known.
9.4 Hypothesis Tests About a Population
Proportion: Large Samples
Often we want to conduct a test of hypothesis about a population proportion. For example, 33%
of the students listed in Who's Who Among American High School Students said that drugs and
alcohol are the most serious problems facing their high schools. A sociologist may want to check
if this percentage still holds. As another example, a mail-order company claims that 90% of all
orders it receives are shipped within 72 hours. The company's management may want to deter-
mine from time to time whether or not this claim is true.
This section presents the procedure to perform tests of hypotheses about the population pro-
portion, p, for large samples. The procedures to make such tests are similar in many respects
to the ones for the population mean, p. Again, the test can be two-tailed or one-tailed. We know
from Chapter 7 that when the sample size is large, the sample proportion, p, is approximately
normally distributed with its mean equal to p and standard deviation equal to \fpqjn. Hence,
we use the normal distribution to perform a test of hypothesis about the population proportion,
p, for a large sample. As was mentioned in Chapters 7 and 8, in the case of a proportion, the
sample size is considered to be large when np and nq are both greater than 5.
Test Statistic The value of the test statistic z for the sample proportion, p, is computed as
P-P .
z = where cr& =
a P
The value of p that is used in this formula is the one from the null hypothesis. The value of q
is equal to 1 — p.
The value of z calculated for p using the above formula is also called the observed value of z.
In Sections 9.2 and 9.3, we discussed two procedures, the p-value approach and the critical-
value approach, to test hypotheses about p. Here too we will use these two procedures to test
hypotheses about p. The steps used in these procedures are the same as in Sections 9.2 and 9.3.
The only difference is that we will be making tests of hypotheses about p rather than about p.
1. The p-Value Approach
To use the p-value approach to perform a test of hypothesis about p, we will use the same four
steps that we used in such a procedure in Section 9.2. Although the p-value for a test of hy-
pothesis about p can be obtained very easily by using technology, we can use Table IV of Ap-
pendix C to find this /7-value when technology is not available.
Examples 9-9 and 9-10 illustrate the p-value procedure to test a hypothesis about p for a
large sample.
EXAMPLE 9-9
Finding a p-value and
making a decision for a
two-tailed test of hypothesis
about p: large sample.
According to a Nationwide Mutual Insurance Company Driving While Distracted Survey con-
ducted in 2008, 81% of the drivers interviewed said that they have talked on their cell phones
while driving {The New York Times, July 19, 2009). The survey included drivers aged 16 to
61 years selected from 48 states. Assume that this result holds true for the 2008 population of
all such drivers in the United States. In a recent random sample of 1600 drivers aged 16 to 61
years selected from the United States, 83% said that they have talked on their cell phones
while driving. Find the p-value to test the hypothesis that the current percentage of such
9.4 Hypothesis Tests About a Population Proportion: Large Samples 415
drivers who have talked on their cell phones while driving is different from 81%. What is your
conclusion if the significance level is 5%?
Solution Let p be the current proportion of all U.S. drivers aged 16 to 61 years who have
talked on their cell phones while driving, and let p be the corresponding sample proportion.
Then, from the given information,
n = 1600, p = .83, and a = .05
In 2008, 81% of U.S. drivers aged 16 to 61 years said that they had talked on their cell phones
while driving. Hence,
p = .81 and q=l-p = l-M = .19
To calculate the p-value and to make a decision, we apply the following four steps.
Step 1. State the null and alternative hypotheses.
The current percentage of all U.S. drivers aged 16 to 61 years who have talked on their cell
phones while driving will not be different from 81% if p =.81, and the current percentage will
be different from 81% if p =£.81. The null and alternative hypotheses are as follows:
H : /? = .81 (The current percentage is not different from 81%.)
Hi. p ¥= .81 (The current percentage is different from 81 %.)
Step 2. Select the distribution to use.
To check whether the sample is large, we calculate the values of np and nq:
np = 1600(.81) = 1296 and nq = 1600(.19) = 304
Since np and nq are both greater than 5, we can conclude that the sample size is large. Con-
sequently, we will use the normal distribution to find the p-value for this test.
Step 3. Calculate the p-value.
The # sign in the alternative hypothesis indicates that the test is two-tailed. The p-value is
equal to twice the area in the tail of the normal distribution curve to the right of z for p = .83.
This p-value is shown in Figure 9.15. To find this p-value, first we find the test statistic z for
p = .83 as follows:
/(.81)(.19)
1600
.00980752
^ From H a
p-p .83 - .81
z = = = 2.04
o> .00980752
Now we find the area to the right of z = 2.04 from the normal distribution table. This area is
1 — .9793 = .0207. Consequently, the p-value is
p-value = 2(.0207) = .0414
416 Chapter 9 Hypothesis Tests About the Mean and Proportion
Finding a p-value and
making a decision for a right-
tailed test of hypothesis about
p: large sample.
Step 4. Make a decision.
Thus, we can state that for any a greater than .0414 we will reject the null hypothesis, and
for any a less than or equal to .0414 we will not reject the null hypothesis. For our example,
a = .05, which is greater than the p-value of .0414. As a result, we reject H and conclude
that the current percentage of all U.S. drivers aged 16 to 61 years who have talked on their
cell phones while driving is different from .81. Consequently, we can state that the difference
between the hypothesized population proportion of .81 and the sample proportion of .83 is too
large to be attributed to sampling error alone when a = .05. I
When working properly, a machine that is used to make chips for calculators does not pro-
duce more than 4% defective chips. Whenever the machine produces more than 4% defective
chips, it needs an adjustment. To check if the machine is working properly, the quality con-
trol department at the company often takes samples of chips and inspects them to determine
if they are good or defective. One such random sample of 200 chips taken recently from the
production line contained 12 defective chips. Find the p-value to test the hypothesis whether
or not the machine needs an adjustment. What would your conclusion be if the significance
level is 2.5%?
Solution Let p be the proportion of defective chips in all chips produced by this machine,
and let p be the corresponding sample proportion. Then, from the given information,
When the machine is working properly, it does not produce more than 4% defective chips.
Hence, assuming that the machine is working properly, we obtain
To calculate the /7-value and to make a decision, we apply the following four steps.
Step 1. State the null and alternative hypotheses.
The machine will not need an adjustment if the percentage of defective chips is 4% or less,
and it will need an adjustment if this percentage is greater than 4%. Hence, the null and al-
ternative hypotheses are as follows:
Since np and nq are both greater than 5, we can conclude that the sample size is large. Con-
sequently, we will use the normal distribution to find the p-value for this test.
Step 3. Calculate the p-value.
The > sign in the alternative hypothesis indicates that the test is right-tailed. The p-value
is given by the area in the upper tail of the normal distribution curve to the right of z for
p = .06. This p-value is shown in Figure 9.16. To find this p-value, first we find the test sta-
tistic z for p = .06 as follows:
■ EXAMPLE 9-10
n = 200, p = 12/200 = .06, and a = .025
p = .04 and q = 1 - p = 1 - .04 = .96
H : p < .04 (The machine does not need an adjustment.)
H x : p > .04 (The machine needs an adjustment.)
Step 2. Select the distribution to use.
To check if the sample is large, we calculate the values of np and nq:
np = 200(.04) = 8 and nq = 200(.96) = 192
.06 - .04
From H,
P ~ P
= 1.44
z =
.01385641
9.4 Hypothesis Tests About a Population Proportion: Large Samples 417
Now we find the area to the right of z = 1 -44 from the normal distribution table. This area is
1 — .9251 = .0749. Consequently, the /7-value is
p-value = .0749
Step 4. Make a decision.
Thus, we can state that for any a greater than .0749 we will reject the null hypothesis, and
for any a less than or equal to .0749 we will not reject the null hypothesis. For our example,
a = .025, which is less than the />value of .0749. As a result, we fail to reject H and con-
clude that the machine does not need an adjustment. I
2. The Critical-Value Approach
As mentioned in Section 9.2, this procedure is also called the traditional or classical approach.
In this procedure, we have a predetermined value of the significance level a. The value of a
gives the total area of the rejection region(s). First we find the critical value(s) of z from the
normal distribution table for the given significance level. Then we find the value of the test sta-
tistic z for the observed value of the sample statistic p. Finally we compare these two values
and make a decision. Remember, if the test is one-tailed, there is only one critical value of z,
and it is obtained by using the value of a, which gives the area in the left or right tail of the
normal distribution curve, depending on whether the test is left-tailed or right-tailed, respec-
tively. However, if the test is two-tailed, there are two critical values of z, and they are obtained
by using a/2 area in each tail of the normal distribution curve. The value of the test statistic z
is obtained as mentioned earlier in this section.
Examples 9-11 and 9-12 describe the procedure to test a hypothesis about p using the
critical-value approach and the normal distribution.
EXAMPLE 9-11
Refer to Example 9-9. According to a Nationwide Mutual Insurance Company Driving While
Distracted Survey conducted in 2008, 81% of the drivers interviewed said that they have talked
on their cell phones while driving (The New York Times, July 19, 2009). The survey included driv-
ers aged 16 to 61 years selected from 48 states. Assume that this result holds true for the 2008
population of all such drivers in the United States. In a recent random sample of 1600 drivers
aged 16 to 61 years selected from the United States, 83% said that they have talked on their cell
phones while driving. Using the 5% significance level, can you conclude that the current percent-
age of such drivers who have talked on their cell phones while driving is different from 81%?
Making a two-tailed test
of hypothesis about p using
the critical-value approach:
large sample.
Solution Let p be the current proportion of all U.S. drivers aged 16 to 61 years who have
talked on their cell phones while driving, and let p be the corresponding sample proportion.
Then, from the given information,
n = 1600, p = .83, and a = .05
In 2008, 81% of U.S. drivers aged 16 to 61 years said that they had talked on their cell phones
while driving. Hence,
p = .81 and q = l-p = l~M = .19
To use the critical-value approach to make a decision, we apply the following five steps.
418 Chapter 9 Hypothesis Tests About the Mean and Proportion
Step 1. State the null and alternative hypotheses.
The current percentage of all U.S. drivers aged 16 to 61 years who have talked on their
cell phones while driving will not be different from 81% if p =.81, and the current per-
centage will be different from 81% if p #.81. The null and alternative hypotheses are as
follows:
H : p = .81 (The current percentage is not different from 81%.)
H x : p # .81 (The current percentage is different from 81%.)
Step 2. Select the distribution to use.
To check whether the sample is large, we calculate the values of tip and nq:
np = 1600(.81) = 1296 and nq = 1600(.19) = 304
Since np and nq are both greater than 5, we can conclude that the sample size is large. Con-
sequently, we will use the normal distribution to make the test.
Step 3. Determine the rejection and nonrejection regions.
The ¥= sign in the alternative hypothesis indicates that the test is two-tailed. The signifi-
cance level is .05. Therefore, the total area of the two rejection regions is .05, and the rejec-
tion region in each tail of the sampling distribution of p is a/2 = .05/2 = .025. The critical
values of z, obtained from the standard normal distribution table, are — 1.96 and 1.96, as shown
in Figure 9.17.
Step 4. Calculate the value of the test statistic.
The value of the test statistic z for p = .83 is calculated as follows:
lp~q /(.81)(.19)
a s = ,/ — = ,/- — - — - = .00980752
" V n V 1600
^ From H
p-p .83 - .81
z = ~ = = 2.04
a p .00980752
Step 5. Make a decision.
The value of the test statistic z = 2.04 for p falls in the rejection region. As a result, we
reject // and conclude that the current percentage of all U.S. drivers aged 16 to 61 years who
have talked on their cell phones while driving is different from .81. Consequently, we can state
9.4 Hypothesis Tests About a Population Proportion: Large Samples 419
that the difference between the hypothesized population proportion of .81 and the sample pro-
portion of .83 is too large to be attributed to sampling error alone when a = .05. H
■ EXAMPLE 9-12
Direct Mailing Company sells computers and computer parts by mail. The company claims
that at least 90% of all orders are mailed within 72 hours after they are received. The quality
control department at the company often takes samples to check if this claim is valid. A re-
cently taken sample of 150 orders showed that 129 of them were mailed within 72 hours. Do
you think the company's claim is true? Use a 2.5% significance level.
Solution Let p be the proportion of all orders that are mailed by the company within
72 hours, and let p be the corresponding sample proportion. Then, from the given
information,
n = 150, p = 129/150 = .86, and a = .025
The company claims that at least 90% of all orders are mailed within 72 hours. Assuming that
this claim is true, the values of p and q are
p = .90 and q=\-p=l-.90 = .10
Step 1. State the null and alternative hypotheses.
The null and alternative hypotheses are
H : p > .90 (The company's claim is true.)
Hi', p < .90 (The company's claim is false.)
Step 2. Select the distribution to use.
We first check whether np and nq are both greater than 5:
np = 150(.90) = 135 > 5 and nq = 150(.10) = 15 > 5
Consequently, the sample size is large. Therefore, we use the normal distribution to make the
hypothesis test about p.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .025. The < sign in the alternative hypothesis indicates that the
test is left-tailed, and the rejection region lies in the left tail of the sampling distribution of p
with its area equal to .025. As shown in Figure 9.18, the critical value of z, obtained from the
normal distribution table for .0250 area in the left tail, is — 1 .96.
Conducting a left-tailed test
of hypothesis about p using the
critical-value approach: large
sample.
FAVORITE
SEAT IN THE
PLANE
Source: The chart reproduced with
permission from USA TODAY, August
27, 2009. Copyright © 2009, USA
TODAY.
USA TODAY Snapshots®
Which seat is your favorite
when you fly?
Window
Middle ) C Aisle )
61%
1%
n
I
38%
1
roe: 3M Privacy
*
*
rs survey of 806
■ *
lis 18 and older.
By Jje Yang and Davf Merrill. USA TODAY
The accompanying chart shows the percentage of people who prefer specific seats in the plane when they
fly. As mentioned in the chart, it shows that 61% of the adults prefer a window seat, 38% prefer an aisle
seat, and only 1% prefer the middle seat. Note that these results are based on a sample of 806 adults.
Suppose that these results were true for the population of such adults at the time of the survey, and
that we want to check if the current percentage of all adults who prefer the window seat when they fly is
still 61%. Suppose we take a random sample of 1000 adults and ask them which seat is their favorite when
they fly. Of them, 640 say that they prefer a window seat. Suppose the significance level is 1%. The test is
two-tailed. The null and alternative hypotheses are
H :p
.61
and
H,:p * .61
Here, n = 1 000, p = .640/ 1 000 = .64, a = .01 , and a/2 = .005. The sample is large. (The reader should check
that np and nq are both greater than 5.) Using the normal distribution to make the test, the critical values of
z for .0050 and .9950 areas to the left are -2.58 and 2.58. We find the observed value of z as follows.
(-61X-39)
1000
.01542401
P-P
.64 - .61
.01542401
1.95
The value of the test statistic z = 1 .95 for p is between the two critical values of z = -2.58 and z =
2.58, and it falls in the nonrejection region. Consequently, we fail to reject H and conclude that the cur-
rent percentage of adults who prefer the window seat when they fly is not significantly different from 61%.
We can use the p-value approach too. From the normal distribution table, the area under the normal
curve to the right of z = 1.95 is 1 -.9744 = .0256. Therefore, the p-value is 2(.0256) =.0512. Since a =
.01 is smaller than .0512, we fail to reject the null hypothesis.
Step 4. Calculate the value of the test statistic.
The value of the test statistic z for p = .86 is calculated as follows:
.90 .10
- — = .02449490
150
^ From H
- 86 "- 90 -1.63
.02449490
420
9.4 Hypothesis Tests About a Population Proportion: Large Samples
421
Step 5. Make a decision.
The value of the test statistic z = —1.63 is greater than the critical value of z — —1.96,
and it falls in the nonrejection region. Therefore, we fail to reject H . We can state that the
difference between the sample proportion and the hypothesized value of the population pro-
portion is small, and this difference may have occurred owing to chance alone. Therefore, the
proportion of all orders that are mailed within 72 hours is at least 90%, and the company's
claim is true.
EXERCISES
CONCEPTS AND PROCEDURES
9.74 Explain when a sample is large enough to use the normal distribution to make a test of hypothesis
about the population proportion.
9.75 In each of the following cases, do you think the sample size is large enough to use the normal dis-
tribution to make a test of hypothesis about the population proportion? Explain why or why not.
a.
;/
= 40
and
P =
.11
b.
n
= 100
and
P =
.73
c.
n
= 80
and
P =
.05
d.
n
= 50
and
P =
.14
9.76 In each of the following cases, do you think the sample size is large enough to use the normal dis-
tribution to make a test of hypothesis about the population proportion? Explain why or why not.
a. n = 30 and p = .65
b. n = 70 and p = .05
c. n = 60 and p = .06
d. n = 900 and p = .17
9.77 For each of the following examples of tests of hypothesis about the population proportion, show the
rejection and nonrejection regions on the graph of the sampling distribution of the sample proportion.
a. A two-tailed test with a = .10
b. A left-tailed test with a = .01
c. A right-tailed test with a = .05
9.78 For each of the following examples of tests of hypothesis about the population proportion, show the
rejection and nonrejection regions on the graph of the sampling distribution of the sample proportion.
a. A two-tailed test with a = .05
b. A left-tailed test with a = .02
c. A right-tailed test with a = .025
9.79 A random sample of 500 observations produced a sample proportion equal to .38. Find the critical
and observed values of z for each of the following tests of hypotheses using a = .05.
a. H : p = .30 versus H x : p > .30
b. H : p = .30 versus H^. p + .30
9.80 A random sample of 200 observations produced a sample proportion equal to .60. Find the critical
and observed values of z for each of the following tests of hypotheses using a = .01.
a. H : p = .63 versus H x : p < .63
b. H : p = .63 versus H{. p + .63
9.81 Consider the null hypothesis H : p = .65. Suppose a random sample of 1000 observations is taken
to perform this test about the population proportion. Using a = .05, show the rejection and nonrejection
regions and find the critical value(s) of z for a
a. left-tailed test
b. two-tailed test
c. right-tailed test
9.82 Consider the null hypothesis H a : p = .25. Suppose a random sample of 400 observations is taken to
perform this test about the population proportion. Using a = .01, show the rejection and nonrejection re-
gions and find the critical value(s) of z for a
a. left-tailed test
b. two-tailed test
c. right- tailed test
422 Chapter 9 Hypothesis Tests About the Mean and Proportion
9.83 Consider H : p = .70 versus H x :p + .70.
a. A random sample of 600 observations produced a sample proportion equal to .68. Using a = .01,
would you reject the null hypothesis?
b. Another random sample of 600 observations taken from the same population produced a sample
proportion equal to .76. Using a = .01, would you reject the null hypothesis?
Comment on the results of parts a and b.
9.84 Consider H : p = .45 versus H{. p < .45.
a. A random sample of 400 observations produced a sample proportion equal to .42. Using a = .025,
would you reject the null hypothesis?
b. Another random sample of 400 observations taken from the same population produced a sample
proportion of .39. Using a = .025, would you reject the null hypothesis?
Comment on the results of parts a and b.
9.85 Make the following hypothesis tests about p.
a. H : p = .45, H x : p =h .45, n = 100, p = .49, a = .10
b. H : p = .72, H l :p< .72, n = 700, p = .64, a = .05
c. H : p = .30, Hi: p > .30, n = 200, p = .33, a = .01
9.86 Make the following hypothesis tests about p.
a. H : p = .57, Hi. p + .57, n = 800, p = .50, a = .05
b. H : p = .26, Hi. p < .26, n = 400, p = .23, a = .01
c. H : p = .84, Hi. p > .84, n = 250, p = .85, a = .025
■ APPLICATIONS
9.87 According to a 2008 survey by the Royal Society of Chemistry, 30% of adults in Great Britain stated
that they typically run the water for a period of 6 to 10 minutes while taking the shower (http://www.rsc.org/
AboutUs/News/PressReleases/2008/EuropeanShowerHabits.asp). Suppose that in a recent survey of 400
adults in Great Britain, 104 stated that they typically run the water for a period of 6 to 10 minutes when
they take a shower. At the 5% significance level, can you conclude that the proportion of all adults in
Great Britain who typically run the water for a period of 6 to 10 minutes when they take a shower is less
than .30? Use both the p-value and the critical value approaches.
9.88 In a 2009 nonscientific poll on the Web site of the Daily Gazette of Schenectady, New York, read-
ers were asked the following question: "Are you less inclined to buy a General Motors or Chrysler vehi-
cle now that they have filed for bankruptcy?" Of the respondents, 56.1% answered "Yes" (http://www.
dailygazette.com/polls/2009/jun/Bankruptcy/). In a recent survey of 1200 adult Americans who were asked
the same question, 615 answered "Yes." Can you reject the null hypothesis at the 1% significance level in
favor of the alternative that the percentage of all adult Americans who are less inclined to buy a General
Motors or Chrysler vehicle since the companies filed for bankruptcy is different from 56.1%? Use both
the p-value and the critical- value approaches.
9.89 In a 2009 nonscientific poll on www.ESPN.com, 67% of the respondents believed that Roger Fed-
erer was going to defeat Andy Roddick in the 2009 Wimbledon Gentlemen's singles championship. Sup-
pose that a survey of 150 tennis fans conducted in Europe at the same time resulted in 118 who believed
that Federer was going to win. Perform a hypothesis test to determine if it is reasonable to conclude that
the percentage of all European tennis fans who believed that Federer was going to win the 2009 champi-
onship was higher than 67%, the result in the ESPN.com poll. Use a 2% significance level and both the
p-value and the critical-value approaches.
9.90 As reported on carefair.com (November 15, 2006), 40% of women aged 30 years and older would
rather get Botox injections than spend a week in Paris. In a recent survey of 400 women aged 65 years
and older, 108 women would rather get Botox injections than spend a week in Paris. Using a 10% signif-
icance level, perform a test of hypothesis to determine whether the current percentage of women aged 65
years or older who would rather get Botox injections than spend a week in Paris is less than 40%. Use
both the p-value and the critical-value approaches.
9.91 According to the American Diabetes Association (www.diabetes.org), 23.1% of Americans aged 60
years or older had diabetes in 2007. A recent random sample of 200 Americans aged 60 years or older
showed that 52 of them have diabetes. Using a 5% significance level, perform a test of hypothesis to de-
termine if the current percentage of Americans aged 60 years or older who have diabetes is higher than
that in 2007. Use both the p-value and the critical-value approaches.
9.4 Hypothesis Tests About a Population Proportion: Large Samples
423
9.92 As noted in U.S. Senate Resolution 28, 9.3% of Americans speak their native language and another
language fluently (Source: www.actfl.org/i4a/pages/index. cfm?pageid= 3782). Suppose that in a recent
sample of 880 Americans, 69 speak their native language and another language fluently. Is there signifi-
cant evidence at the 10% significance level that the percentage of all Americans who speak their native
language and another language fluently is different from 9.3%? Use both the p-value and the critical-value
approaches.
9.93 A food company is planning to market a new type of frozen yogurt. However, before marketing this
yogurt, the company wants to find what percentage of the people like it. The company's management has
decided that it will market this yogurt only if at least 35% of the people like it. The company's research
department selected a random sample of 400 persons and asked them to taste this yogurt. Of these 400
persons, 112 said they liked it.
a. Testing at the 2.5% significance level, can you conclude that the company should market this
yogurt?
b. What will your decision be in part a if the probability of making a Type I error is zero? Explain.
c. Make the test of part a using the p-value approach and a = .025.
9.94 A mail-order company claims that at least 60% of all orders are mailed within 48 hours. From time
to time the quality control department at the company checks if this promise is fulfilled. Recently the qual-
ity control department at this company took a sample of 400 orders and found that 208 of them were
mailed within 48 hours of the placement of the orders.
a. Testing at the 1% significance level, can you conclude that the company's claim is true?
b. What will your decision be in part a if the probability of making a Type I error is zero? Explain.
c. Make the test of part a using the p-value approach and a = .01.
9.95 Brooklyn Corporation manufactures DVDs. The machine that is used to make these DVDs is known
to produce not more than 5% defective DVDs. The quality control inspector selects a sample of 200 DVDs
each week and inspects them for being good or defective. Using the sample proportion, the quality con-
trol inspector tests the null hypothesis p £ .05 against the alternative hypothesis p > .05, where p is the
proportion of DVDs that are defective. She always uses a 2.5% significance level. If the null hypothesis
is rejected, the production process is stopped to make any necessary adjustments. A recent sample of 200
DVDs contained 17 defective DVDs.
a. Using the 2.5% significance level, would you conclude that the production process should be
stopped to make necessary adjustments?
b. Perform the test of part a using a 1% significance level. Is your decision different from the one
in part a?
Comment on the results of parts a and b.
9.96 Shulman Steel Corporation makes bearings that are supplied to other companies. One of the machines
makes bearings that are supposed to have a diameter of 4 inches. The bearings that have a diameter of
either more or less than 4 inches are considered defective and are discarded. When working properly, the
machine does not produce more than 7% of bearings that are defective. The quality control inspector selects
a sample of 200 bearings each week and inspects them for the size of their diameters. Using the sample pro-
portion, the quality control inspector tests the null hypothesis p ^ .07 against the alternative hypothesis p >
.07, where p is the proportion of bearings that are defective. He always uses a 2% significance level. If the
null hypothesis is rejected, the machine is stopped to make any necessary adjustments. One sample of 200
bearings taken recently contained 22 defective bearings.
a. Using the 2% significance level, will you conclude that the machine should be stopped to make
necessary adjustments?
b. Perform the test of part a using a 1% significance level. Is your decision different from the one
in part a?
Comment on the results of parts a and b.
*9.97 Two years ago, 75% of the customers of a bank said that they were satisfied with the services pro-
vided by the bank. The manager of the bank wants to know if this percentage of satisfied customers has
changed since then. She assigns this responsibility to you. Briefly explain how you would conduct such
a test.
*9.98 A study claims that 65% of students at all colleges and universities hold off-campus (part-time or
full-time) jobs. You want to check if the percentage of students at your school who hold off-campus jobs
is different from 65%. Briefly explain how you would conduct such a test. Collect data from 40 students
at your school on whether or not they hold off-campus jobs. Then, calculate the proportion of students in
this sample who hold off-campus jobs. Using this information, test the hypothesis. Select your own sig-
nificance level.
424 Chapter 9 Hypothesis Tests About the Mean and Proportion
USES AND MISUSES...
FOLLOW THE RECIPE
Hypothesis testing is one of the most powerful and dangerous tools
of statistics. It allows us to make statements about a population
and attach a degree of uncertainty to these statements. Pick up a
newspaper and flip through it; rare will be the day when the pa-
per does not contain a story featuring a statistical result, often re-
ported with a significance level. Given that the subjects of these
reports— public health, the environment, and so on— are important
to our lives, it is critical that we perform the statistical calculations
and interpretations properly. The first step, one that you should
look for when reading statistical results, is proper formulation/
specification.
Formulation or specification, simply put, is the list of steps you
perform when constructing a hypothesis test. In this chapter, these
steps are: stating the null and alternative hypotheses; selecting the
appropriate distribution; and determining the rejection and nonrejec-
tion regions. Once these steps are performed, all you need to do is
to calculate the p-value or the test statistic to complete the hypothe-
sis test. It is important to beware of traps in the specification.
Though it might seem obvious, stating the hypothesis properly
can be difficult. For hypotheses around a population mean, the null
and alternative hypotheses are mathematical statements that do not
overlap and also provide no holes. Suppose that a confectioner states
that the average mass of his chocolate bars is 100 grams. The null
hypothesis is that the mass of the bars is 100 grams, and the alter-
native hypothesis is that the mass of the bars is not 1 00 grams. When
you take a sample of chocolate bars and measure their masses, all
possibilities for the sample mean will fall within one of your deci-
sion regions. The problem is a little more difficult for hypotheses
based on proportions. Make sure that you only have two categories.
For example, if you are trying to determine the percentage of the
population that has blonde hair, your groups are "blonde" and "not
blonde." You need to decide how to categorize bald people before
you conduct this experiment: Do not include bald people in the
survey.
Finally, beware of numerical precision. When your sample is large
and you assume that it has a normal distribution, the rejection re-
gion for a two-tailed test using the normal distribution with a signif-
icance level of 5% will be values of the sample mean that are far-
ther than 1.96 standard deviations from the assumed mean. When
you perform your calculations, the sample mean may fall on the bor-
der of your decision region. Remember that there is measurement
error and sample error that you cannot account for. In this case, it is
probably best to adjust your significance level so that the sample
mean falls squarely in a decision region.
SHOULD WE TOSS OUT THE FLIP?
Who goes first? Who gets to pick the first piece of cake? Who gets
the ball first? Who gets to sit in front? Check out playgrounds,
kitchens, and driveways full of children and exasperated parents try-
ing to obtain peace by offering a "fair" decision-making process. The
short straw. Rock, Paper, Scissors. And, of course, the coin flip. For
years we have known that the coin flip is a fair, 50-50 method of
decision making. Recent research is questioning what we have
known.
A recent study by researchers at Stanford University has exam-
ined the laws of mechanics and their impact on the outcome of coin
flips. According to the results of the study, "for natural flips, the chance
of a coin coming up on the same side as it started is about 51 per-
cent. Heads facing up predicts heads; tails facing up predicts tails." In
other words, when you place the coin on your thumb to flip it, if
head is facing up, the coin is more likely to land with head up. Sim-
ilarly, if tail is facing up, the coin is more likely to land with tail up.
But before anyone suggests that the National Football League review
all game outcomes for biased coins, there are some important points
to consider.
As David Adler pointed out in his article, the Stanford team
"built a coin-tossing machine and filmed it using a slow-motion
camera. This confirmed that the outcome of flips isn't random. The
machine could make the toss come out heads every time." David
Adler quickly notes that the coin flips that occur in society, espe-
cially in football games, are subject to a number of additional fac-
tors that a machine in a climate-controlled room would not have
to deal with. First, humans are not able to apply the same amount
of force to a coin on every flip, which impacts the dynamics of a
flip. Second, the flips at the beginning of football games land on
the playing surface, which could be dry, frozen, muddy, grassy, or
artificial turf, all of which can impact the rebound of the coin when
it hits the surface. In addition, as fans of certain teams such as the
Chicago Bears, Green Bay Packers, New York Giants, and New Eng-
land Patriots are well aware, strong winds can also impact the flight
of the coin.
When asked whether the coin flip should be replaced as the
method for starting a football game, one of the members of the
Stanford research team replied, "there is no reason to change
the way the coin flip is done, as long as the person calling the flip
doesn't know how the coin is going to start out. In football, the
tosser is never the caller; the tosser is supposed to be a referee.
But if you are both the caller and the tosser... well, that changes
things. Knowing about the bias in coin tosses give you an edge,
albeit a tiny one."
Note: Mini-Project 9-4 will provide you with an opportunity to test
the conclusion given in this Uses and Misuses section.
Source: David E. Adler: Flipping Out— Think a coin toss has a 50-50
chance? Think again. July 28, 2009. http://www.thebigmoney.com/
articles/hey-wait-minute/2009/07/28/flipping-out?page=0,0&g=l
Supplementary Exercises 425
Glossary
a The significance level of a test of hypothesis that denotes the
probability of rejecting a null hypothesis when it actually is true.
(The probability of committing a Type I error.)
Alternative hypothesis A claim about a population parameter that
will be true if the null hypothesis is false.
/3 The probability of not rejecting a null hypothesis when it actu-
ally is false. (The probability of committing a Type II error.)
Critical value or critical point One or two values that divide the
whole region under the sampling distribution of a sample statistic
into rejection and nonrejection regions.
Left-tailed test A test in which the rejection region lies in the left
tail of the distribution curve.
Null hypothesis A claim about a population parameter that is as-
sumed to be true until proven otherwise.
Observed value of z or t The value of z or t calculated for a sam-
ple statistic such as the sample mean or the sample proportion.
One-tailed test A test in which there is only one rejection region,
either in the left tail or in the right tail of the distribution curve.
/7-value The smallest significance level at which a null hypothesis
can be rejected.
Right-tailed test A test in which the rejection region lies in the
right tail of the distribution curve.
Significance level The value of a that gives the probability of com-
mitting a Type I error.
Test statistic The value of z or t calculated for a sample statistic
such as the sample mean or the sample proportion.
Two-tailed test A test in which there are two rejection regions,
one in each tail of the distribution curve.
Type I error An error that occurs when a true null hypothesis is
rejected.
Type II error An error that occurs when a false null hypothesis
is not rejected.
Supplementary Exercises
9.99 Consider the following null and alternative hypotheses:
H : p. = 120 versus H t : p > 120
A random sample of 81 observations taken from this population produced a sample mean of 123.5. The
population standard deviation is known to be 15.
a. If this test is made at the 2.5% significance level, would you reject the null hypothesis? Use
the critical-value approach.
b. What is the probability of making a Type I error in part a?
c. Calculate the p-value for the test. Based on this p-value, would you reject the null hypothesis
if a = .01? What if a = .05?
9.100 Consider the following null and alternative hypotheses:
H : p = 40 versus H^, p + 40
A random sample of 64 observations taken from this population produced a sample mean of 38.4. The
population standard deviation is known to be 6.
a. If this test is made at the 2% significance level, would you reject the null hypothesis? Use the
critical-value approach.
b. What is the probability of making a Type I error in part a?
c. Calculate the p-value for the test. Based on this p-value, would you reject the null hypothesis
if a = .01? What if a = .05?
9.101 Consider the following null and alternative hypotheses:
H : p = .82 versus H^. p + .82
A random sample of 600 observations taken from this population produced a sample proportion of .86.
a. If this test is made at the 2% significance level, would you reject the null hypothesis? Use the
critical-value approach.
b. What is the probability of making a Type I error in part a?
c. Calculate the p-value for the test. Based on this p-value, would you reject the null hypothesis
if a = .025? What if a = .005?
9.102 Consider the following null and alternative hypotheses:
H : p = .44 versus H,: p < .44
426
Chapter 9 Hypothesis Tests About the Mean and Proportion
A random sample of 450 observations taken from this population produced a sample proportion of .39.
a. If this test is made at the 2% significance level, would you reject the null hypothesis? Use the
critical-value approach.
b. What is the probability of making a Type I error in part a?
c. Calculate the p-value for the test. Based on this p-value, would you reject the null hypothesis
if a = .01? What if a = .025?
9.103 According to the most recent Bureau of Labor Statistics survey on time use in the United States,
the average U.S. man spends 67.20 minutes per day eating and drinking. Suppose that a survey of 43
Norwegian men resulted in an average of 81.10 minutes per day eating and drinking [Note: This value is
consistent with the data in a report by the Organization for Economic Cooperation and Development
(Source: http://economix.blogs.nytimes.com/2009/05/05/obesity-and-the-fastness-of-food/)] . Assume that
the population standard deviation for all Norwegian men is 18.30 minutes.
a. Find the p-value for the test of hypothesis with the alternative hypothesis that the average
daily time spent eating and drinking by all Norwegian men is higher than 67.20 minutes.
What is your conclusion at a = .05?
b. Test the hypothesis of part a using the critical-value approach. Use a = .01.
9.104 The mean consumption of water per household in a city was 1245 cubic feet per month. Due to a
water shortage because of a drought, the city council campaigned for water use conservation by house-
holds. A few months after the campaign was started, the mean consumption of water for a sample of 100
households was found to be 1 175 cubic feet per month. The population standard deviation is given to be
250 cubic feet.
a. Find the p-value for the hypothesis test that the mean consumption of water per household has
decreased due to the campaign by the city council. Would you reject the null hypothesis at
a = .025?
b. Make the test of part a using the critical- value approach and a = .025.
9.105 A highway construction zone has a posted speed limit of 40 miles per hour. Workers working at
the site claim that the mean speed of vehicles passing through this construction zone is at least 50 miles
per hour. A random sample of 36 vehicles passing through this zone produced a mean speed of 48 miles
per hour. The population standard deviation is known to be 4 miles per hour.
a. Do you think the sample information is consistent with the workers' claim? Use a = .025.
b. What is the Type I error in this case? Explain. What is the probability of making this error?
c. Will your conclusion of part a change if the probability of making a Type I error is zero?
d. Find the p-value for the test of part a. What is your decision if a = .025?
9.106 In a 2005 Energy Information Administration report (http://www.eia.doe.gov/emeu/reps/enduse/
erOl us.html), the average U.S. household uses 10,654 kilowatt-hours of electricity per year. A random
sample of 85 houses built in the last 12 to 24 months showed that they had an average electricity usage
of 10,278 kilowatt-hours per year. Assume that the population standard deviation is 1576 kilowatt-hours
per year.
a. Using the critical-value approach, can you conclude that the average annual electricity usage of
all houses built in the last 12 to 24 months is less than 10,654 kilowatt-hours? Use a = .01.
b. What is the Type I error in part a? Explain. What is the probability of making this error in
part a?
c. Will your conclusion in part a change if the probability of making a Type I error is zero?
d. Calculate the p-value for the test of part a. What is your conclusion if a = .01?
9.107 A real estate agent claims that the mean living area of all single-family homes in his county is at
most 2400 square feet. A random sample of 50 such homes selected from this county produced the mean
living area of 2540 square feet and a standard deviation of 472 square feet.
a. Using a = .05, can you conclude that the real estate agent's claim is true?
b. What will your conclusion be if a = .01?
Comment on the results of parts a and b.
9.108 According to an article on PCMag.com, Facebook users spend an average of 190 minutes per
month checking and updating their Facebook page (Source: http://www.pcmag.com/article2/
0,2817,2342757,00.asp). A random sample of 55 college students aged 18 to 22 years with Facebook
accounts resulted in a sample mean time and sample standard deviation of 219.50 and 69.30 minutes
per month, respectively.
a. Using a = .025, can you conclude that the average time spent per month checking and updat-
ing their Facebook pages by all college students aged 18 to 22 years who have Facebook ac-
counts is higher than 190 minutes? Use the critical value approach.
b. Find the range of the p-value for the test of part a. What is your conclusion with a = .025?
Supplementary Exercises 427
9.109 Customers often complain about long waiting times at restaurants before the food is served. A
restaurant claims that it serves food to its customers, on average, within 15 minutes after the order is
placed. A local newspaper journalist wanted to check if the restaurant's claim is true. A sample of 36
customers showed that the mean time taken to serve food to them was 15.75 minutes with a standard de-
viation of 2.4 minutes. Using the sample mean, the journalist says that the restaurant's claim is false. Do
you think the journalist's conclusion is fair to the restaurant? Use the 1% significance level to answer
this question.
9.110 The customers at a bank complained about long lines and the time they had to spend waiting for
service. It is known that the customers at this bank had to wait 8 minutes, on average, before being served.
The management made some changes to reduce the waiting time for its customers. A sample of 60 cus-
tomers taken after these changes were made produced a mean waiting time of 7.5 minutes with a standard
deviation of 2.1 minutes. Using this sample mean, the bank manager displayed a huge banner inside the
bank mentioning that the mean waiting time for customers has been reduced by new changes. Do you
think the bank manager's claim is justifiable? Use the 2.5% significance level to answer this question. Use
both approaches.
9.111 The administrative office of a hospital claims that the mean waiting time for patients to get treat-
ment in its emergency ward is 25 minutes. A random sample of 16 patients who received treatment in the
emergency ward of this hospital produced a mean waiting time of 27.5 minutes with a standard deviation
of 4.8 minutes. Using the 1% significance level, test whether the mean waiting time at the emergency ward
is different from 25 minutes. Assume that the waiting times for all patients at this emergency ward have
a normal distribution.
9.112 An earlier study claimed that U.S. adults spent an average of 114 minutes with their families per
day. A recently taken sample of 25 adults from a city showed that they spend an average of 109 minutes
per day with their families. The sample standard deviation is 1 1 minutes. Assume that the times spent by
adults with their families have an approximately normal distribution.
a. Using the 1% significance level, test whether the mean time spent currently by all adults with
their families in this city is different from 1 14 minutes a day.
b. Suppose the probability of making a Type I error is zero. Can you make a decision for the test
of part a without going through the five steps of hypothesis testing? If yes, what is your deci-
sion? Explain.
9.113 A computer company that recently introduced a new software product claims that the mean time it
takes to learn how to use this software is not more than 2 hours for people who are somewhat familiar
with computers. A random sample of 12 such persons was selected. The following data give the times
taken (in hours) by these persons to learn how to use this software.
1.75 2.25 2.40 1.90 1.50 2.75
2.15 2.25 1.80 2.20 3.25 2.60
Test at the 1% significance level whether the company's claim is true. Assume that the times taken by all
persons who are somewhat familiar with computers to learn how to use this software are approximately
normally distributed.
9.114 A company claims that its 8-ounce low-fat yogurt cups contain, on average, at most 150 calories
per cup. A consumer agency wanted to check whether or not this claim is true. A random sample of 10
such cups produced the following data on calories.
147 159 153 146 144 161 163 153 143 158
Test at the 2.5% significance level whether the company's claim is true. Assume that the numbers of calo-
ries for such cups of yogurt produced by this company have an approximately normal distribution.
9.115 A 2008 AARP survey reported that 85% of U.S. workers aged 50 years and older with at least one
4-year college degree had taken employer-based training within the previous 2 years, compared to only
50% of workers aged 50 years and older with a high school degree or less. In a current survey of 640 U.S.
workers aged 50 years and older with a high school degree or less, 341 had taken employer-based train-
ing within the previous 2 years.
a. Using the critical-value approach and a = .05, test whether the current percentage of all U.S.
workers aged 50 years and older with a high school degree or less who have taken employer-
based training within the previous 2 years is different from 50%.
b. How do you explain the Type I error in part a? What is the probability of making this error in
part a?
c. Calculate the p- value for the test of part a. What is your conclusion if a = .05?
428 Chapter 9 Hypothesis Tests About the Mean and Proportion
9.116 In an observational study at Turner Field in Atlanta, Georgia, 43% of the men were observed not
washing their hands after going to the bathroom (see Exercise 7.80). Suppose that in a random sample of
95 men who used the bathroom at Camden Yards in Baltimore, Maryland, 26 did not wash their hands.
a. Using the critical- value approach and a = .10, test whether the percentage of all men at
Camden Yards who use the bathroom and do not wash their hands is less than 43%.
b. How do you explain the Type I error in part a? What is the probability of making this error in
part a?
c. Calculate the p- value for the test of part a. What is your conclusion if a = .10?
9.117 More and more people are abandoning national brand products and buying store brand products to
save money. The president of a company that produces national brand coffee claims that 40% of the peo-
ple prefer to buy national brand coffee. A random sample of 700 people who buy coffee showed that 259
of them buy national brand coffee. Using a = .01, can you conclude that the percentage of people who
buy national brand coffee is different from 40%? Use both approaches to make the test.
9.118 A 2008 study performed by careerbuilder.com entitled No, Really, Your Excuse is Totally Believ-
able! notes that 11% of workers who call in sick do so to catch up on housework. Suppose that in a sur-
vey of 675 male workers who have called in sick, 61 did so to have time to catch up on housework. At
the 2% significance level, can you conclude that the proportion of all male workers who call in sick do
so to catch up on housework is different from 11%?
9.119 Mong Corporation makes auto batteries. The company claims that 80% of its LL70 batteries are
good for 70 months or longer. A consumer agency wanted to check if this claim is true. The agency took
a random sample of 40 such batteries and found that 75% of them were good for 70 months or longer.
a. Using the 1% significance level, can you conclude that the company's claim is false?
b. What will your decision be in part a if the probability of making a Type I error is zero? Explain.
9.120 Dartmouth Distribution Warehouse makes deliveries of a large number of products to its customers.
To keep its customers happy and satisfied, the company's policy is to deliver on time at least 90% of all
the orders it receives from its customers. The quality control inspector at the company quite often takes
samples of orders delivered and checks to see whether this policy is maintained. A recent sample of 90
orders taken by this inspector showed that 75 of them were delivered on time.
a. Using the 2% significance level, can you conclude that the company's policy is maintained?
b. What will your decision be in part a if the probability of making a Type I error is zero? Explain.
Advanced Exercises
9.121 Professor Hansen believes that some people have the ability to predict in advance the outcome of a spin
of a roulette wheel. He takes 100 student volunteers to a casino. The roulette wheel has 38 numbers, each of
which is equally likely to occur. Of these 38 numbers, 18 are red, 18 are black, and 2 are green. Each student
is to place a series of five bets, choosing either a red or a black number before each spin of the wheel. Thus,
a student who bets on red has an 18/38 chance of winning that bet. The same is true of betting on black.
a. Assuming random guessing, what is the probability that a particular student will win all five
of his or her bets?
b. Suppose for each student we formulate the hypothesis test
H ; The student is guessing
H{. The student has some predictive ability
Suppose we reject H Q only if the student wins all five bets. What is the significance level?
c. Suppose that 2 of the 100 students win all five of their bets. Professor Hansen says, "For these
two students we can reject H and conclude that we have found two students with some ability
to predict." What do you make of Professor Hansen's conclusion?
9.122 Acme Bicycle Company makes derailleurs for mountain bikes. Usually no more than 4% of these
parts are defective, but occasionally the machines that make them get out of adjustment and the rate of
defectives exceeds 4%. To guard against this, the chief quality control inspector takes a random sample
of 130 derailleurs each week and checks each one for defects. If too many of these parts are defective, the
machines are shut down and adjusted. To decide how many parts must be defective to shut down the ma-
chines, the company's statistician has set up the hypothesis test
H : p < .04 versus H{.p > .04
where p is the proportion of defectives among all derailleurs being made currently. Rejection of H would
call for shutting down the machines. For the inspector's convenience, the statistician would like the rejec-
tion region to have the form, "Reject H if the number of defective parts is C or more." Find the value of
C that will make the significance level (approximately) .05.
Supplementary Exercises 429
9.123 Alpha Airlines claims that only 15% of its flights arrive more than 10 minutes late. Let p be the
proportion of all of Alpha's flights that arrive more than 10 minutes late. Consider the hypothesis test
H : p £ .15 versus H{. p > .15
Suppose we take a random sample of 50 flights by Alpha Airlines and agree to reject H Q if 9 or more of
them arrive late. Find the significance level for this test.
9.124 The standard therapy used to treat a disorder cures 60% of all patients in an average of 140 visits.
A health care provider considers supporting a new therapy regime for the disorder if it is effective in re-
ducing the number of visits while retaining the cure rate of the standard therapy. A study of 200 patients
with the disorder who were treated by the new therapy regime reveals that 108 of them were cured in an
average of 132 visits with a standard deviation of 38 visits. What decision should be made using a .01
level of significance?
9.125 The print on the packages of 100-watt General Electric soft-white lightbulbs states that these
lightbulbs have an average life of 750 hours. Assume that the standard deviation of the lengths of
lives of these lightbulbs is 50 hours. A skeptical consumer does not think these lightbulbs last as long
as the manufacturer claims, and she decides to test 64 randomly selected lightbulbs. She has set up
the decision rule that if the average life of these 64 lightbulbs is less than 735 hours, then she will
conclude that GE has printed too high an average length of life on the packages and will write them
a letter to that effect. Approximately what significance level is the consumer using? Approximately
what significance level is she using if she decides that GE has printed too high an average length of
life on the packages if the average life of the 64 lightbulbs is less than 700 hours? Interpret the val-
ues you get.
9.126 Thirty percent of all people who are inoculated with the current vaccine used to prevent a disease
contract the disease within a year. The developer of a new vaccine that is intended to prevent this disease
wishes to test for significant evidence that the new vaccine is more effective.
a. Determine the appropriate null and alternative hypotheses.
b. The developer decides to study 100 randomly selected people by inoculating them with the
new vaccine. If 84 or more of them do not contract the disease within a year, the developer
will conclude that the new vaccine is superior to the old one. What significance level is the
developer using for the test?
c. Suppose 20 people inoculated with the new vaccine are studied and the new vaccine is con-
cluded to be better than the old one if fewer than 3 people contract the disease within a year.
What is the significance level of the test?
9.127 Since 1984, all automobiles have been manufactured with a middle tail-light. You have been hired
to answer the question, Is the middle tail-light effective in reducing the number of rear-end collisions?
You have available to you any information you could possibly want about all rear-end collisions involv-
ing cars built before 1984. How would you conduct an experiment to answer the question? In your an-
swer, include things like (a) the precise meaning of the unknown parameter you are testing; (b) H and
Hi, (c) a detailed explanation of what sample data you would collect to draw a conclusion; and (d) any
assumptions you would make, particularly about the characteristics of cars built before 1984 versus those
built since 1984.
9.128 Before a championship football game, the referee is given a special commemorative coin to toss to
decide which team will kick the ball first. Two minutes before game time, he receives an anonymous tip
that the captain of one of the teams may have substituted a biased coin that has a 70% chance of show-
ing heads each time it is tossed. The referee has time to toss the coin 10 times to test it. He decides that
if it shows 8 or more heads in 10 tosses, he will reject this coin and replace it with another coin. Let p be
the probability that this coin shows heads when it is tossed once.
a. Formulate the relevant null and alternative hypotheses (in terms of p) for the referee's test.
b. Using the referee's decision rule, find a for this test.
9.129 In Las Vegas and Atlantic City, New Jersey, tests are performed often on the various gaming de-
vices used in casinos. For example, dice are often tested to determine if they are balanced. Suppose you
are assigned the task of testing a die, using a two-tailed test to make sure that the probability of a 2-spot
is 1/6. Using the 5% significance level, determine how many 2-spots you would have to obtain to reject
the null hypothesis when your sample size is
a. 120 b. 1200 c. 12,000
Calculate the value of p for each of these three cases. What can you say about the relationship between
(1) the difference between p and 1/6 that is necessary to reject the null hypothesis and (2) the sample size
as it gets larger?
430 Chapter 9 Hypothesis Tests About the Mean and Proportion
statistician performs the test H ; /j, = 15 versus H x : /x + 15 and finds fhep-value to be .4546.
The statistician performing the test does not tell you the value of the sample mean and the
value of the test statistic. Despite this, you have enough information to determine the pair of
p-values associated with the following alternative hypotheses.
i. H{. /x < 15 ii. Hy ^ > 15
Note that you will need more information to determine which p-value goes with which alterna-
tive. Determine the pair of p-values. Here the value of the sample mean is the same in both
cases.
b. Suppose the statistician tells you that the value of the test statistic is negative. Match the
p-values with the alternative hypotheses.
Note that the result for one of the two alternatives implies that the sample mean is not on
the same side of /jl = 15 as the rejection region. Although we have not discussed this scenario
in the book, it is important to recognize that there are many real-world scenarios in which this
type of situation does occur. For example, suppose the EPA is to test whether or not a com-
pany is exceeding a specific pollution level. If the average discharge level obtained from the
sample falls below the threshold (mentioned in the null hypothesis), then there would be no
need to perform the hypothesis test.
9.131 You read an article that states "50 hypothesis tests of H : /j, = 35 versus Hf. fi + 35 were performed
using a = .05 on 50 different samples taken from the same population with a mean of 35. Of these, 47
tests failed to reject the null hypothesis." Explain why this type of result is not surprising.
Self-Review Test
1. A test of hypothesis is always about
a. a population parameter b. a sample statistic c. a test statistic
2. A Type I error is committed when
a. a null hypothesis is not rejected when it is actually false
b. a null hypothesis is rejected when it is actually true
c. an alternative hypothesis is rejected when it is actually true
3. A Type II error is committed when
a. a null hypothesis is not rejected when it is actually false
b. a null hypothesis is rejected when it is actually true
c. an alternative hypothesis is rejected when it is actually true
4. A critical value is the value
a. calculated from sample data
b. determined from a table (e.g., the normal distribution table or other such tables)
c. neither a nor b
5. The computed value of a test statistic is the value
a. calculated for a sample statistic
b. determined from a table (e.g., the normal distribution table or other such tables)
c. neither a nor b
6. The observed value of a test statistic is the value
a. calculated for a sample statistic
b. determined from a table (e.g., the normal distribution table or other such tables)
c. neither a nor b
7. The significance level, denoted by a, is
a. the probability of committing a Type I error
b. the probability of committing a Type II error
c. neither a nor b
8. The value of j3 gives the
a. probability of committing a Type I error
b. probability of committing a Type II error
c. power of the test
Self-Review Test 43 1
9. The value of 1 — (3 gives the
a. probability of committing a Type I error
b. probability of committing a Type II error
c. power of the test
10. A two-tailed test is a test with
a. two rejection regions b. two nonrejection regions c. two test statistics
11. A one-tailed test
a. has one rejection region b. has one nonrejection region c. both a and b
12. The smallest level of significance at which a null hypothesis is rejected is called
a. a b. p-value c. j3
13. The sign in the alternative hypothesis in a two-tailed test is always
a. < b. > c. +
14. The sign in the alternative hypothesis in a left-tailed test is always
a. < b. > c. +
15. The sign in the alternative hypothesis in a right-tailed test is always
a. < b. > c. +
16. According to http://www.csgnetwork.com/humanh2owater.html, a 175-pound individual who lives in
a warm climate and averages 20 minutes of exercise per day should consume 90.25 ounces of (liquid) wa-
ter per day. Suppose that a random sample of 45 individuals who fall in this category have an average
daily (liquid) water consumption of 83.15 ounces. Assume that the population standard deviation is 15
ounces.
a. Using the critical- value approach and the 1% significance level, can you conclude that the mean
daily (liquid) water consumption by this population differs from the recommended amount of
90.25 ounces?
b. Using the critical-value approach and the 2.5% significance level, can you conclude that the
mean daily (liquid) water consumption by this population is less than the recommended amount
of 90.25 ounces?
c. What is the Type I error in parts a and b? What is the probability of making this error in each
of parts a and b?
d. Calculate the p- value for the test of part a. What is your conclusion if a = .01?
e. Calculate the p- value for the test of part b. What is your conclusion if a = .025?
17. A minor league baseball executive has become concerned about the slow pace of games played in
her league, fearing that it will lower attendance. She meets with the league's managers and umpires and
discusses guidelines for speeding up the games. Before the meeting, the mean duration of nine-inning
games was 3 hours, 5 minutes (i.e., 185 minutes). A random sample of 36 nine-inning games after the
meeting showed a mean of 179 minutes with a standard deviation of 12 minutes.
a. Testing at the 1% significance level, can you conclude that the mean duration of nine-inning
games has decreased after the meeting?
b. What is the Type I error in part a? What is the probability of making this error?
c. What will your decision be in part a if the probability of making a Type I error is zero?
Explain.
d. Find the range for the p-value for the test of part a. What is your decision based on this p-value?
18. An editor of a New York publishing company claims that the mean time taken to write a textbook is
at least 31 months. A sample of 16 textbook authors found that the mean time taken by them to write a
textbook was 25 months with a standard deviation of 7.2 months.
a. Using the 2.5% significance level, would you conclude that the editor's claim is true? Assume
that the time taken to write a textbook is normally distributed for all textbook authors.
b. What is the Type I error in part a? What is the probability of making this error?
c. What will your decision be in part a if the probability of making a Type I error is .001?
19. A financial advisor claims that less then 50% of adults in the United States have a will. A random
sample of 1000 adults showed that 450 of them have a will.
a. At the 5% significance level, can you conclude that the percentage of people who have a will is
less than 50%?
b. What is the Type I error in part a? What is the probability of making this error?
c. What would your decision be in part a if the probability of making a Type I error were zero?
Explain.
d. Find the p-value for the test of hypothesis mentioned in part a. Using this p-value, will you
reject the null hypothesis if a = .05? What if a = .01?
432
Chapter 9 Hypothesis Tests About the Mean and Proportion
Mini-Projects
■ MINI-PROJECT 9-1
According to the information obtained from www.nba.com, the mean height of players who were on the
rosters of National Basketball Association teams during the 2006-2007 season was 78.93 inches. Let /j,
denote the mean height of NBA players for the 2008-2009 season.
a. Take a random sample of 15 players from Dataset 3, the NBA data file that accompanies this text.
Test H Q : /jl = 78.93 inches against H x : /x + 78.93 inches using a = .05. Assume that the popula-
tion of heights is approximately normal.
b. Repeat part a for samples of 3 1 and 45 players, respectively.
c. Did any of the three tests in parts a and b lead to the conclusion that the mean height of NBA
players in 2008-2009 is different from that in 2006-2007?
A thumbtack that is tossed on a desk can land in one of the two ways shown in the following illustration.
Brad and Dan cannot agree on the likelihood of obtaining a head or a tail. Brad argues that obtaining a
tail is more likely than obtaining a head because of the shape of the tack. If the tack had no point at all,
it would resemble a coin that has the same probability of coming up heads or tails when tossed. But the
longer the point, the less likely it is that the tack will stand up on its head when tossed. Dan believes that
as the tack lands tails, the point causes the tack to jump around and come to rest in the heads position.
Brad and Dan need you to settle their dispute. Do you think the tack is equally likely to land heads or
tails? To investigate this question, find an ordinary thumbtack and toss it a large number of times (say,
100 times).
a. What is the meaning, in words, of the unknown parameter in this problem?
b. Set up the null and alternative hypotheses and compute the p-value based on your results from
tossing the tack.
c. How would you answer the original question now? If you decide the tack is not fair, do you side
with Brad or Dan?
d. What would you estimate the value of the parameter in part a to be? Find a 90% confidence in-
terval for this parameter.
e. After doing this experiment, do you think 100 tosses are enough to infer the nature of your tack?
Using your result as a preliminary estimate, determine how many tosses would be necessary to be
95% certain of having 4% accuracy; that is, the margin of error of estimate is ±4%. Have you ob-
served enough tosses?
Collect pennies in the amount of $5. Do not obtain rolls of pennies from a bank because many such rolls
will consist solely of new pennies. Treat these 500 pennies as your population. Determine the ages, in
years, of all these pennies. Calculate the mean and standard deviation of these ages and denote them by
/x and <x, respectively.
a. Take a random sample of 10 pennies from these 500. Find the average age of these 10 pennies,
which is the value of x. Perform a test with the null hypothesis that ti is equal to the value ob-
tained for all 500 pennies and the alternative hypothesis that ti is not equal to this value. Use a
significance level of. 10.
b. Suppose you repeat the procedure of part a nine more times. How many times would you expect
to reject the null hypothesis? Now actually repeat the procedure of part a nine more times, mak-
ing sure that you put the 10 pennies selected each time back in the population and that you mix
all pennies well before taking a sample. How many times did you reject the null hypothesis? Note
■ MINI-PROJECT 9-2
Heads
Tails
■ MINI-PROJECT 9-3
Mini-Projects 433
that you can enter the ages of these 500 pennies in a technology and then use that technology to
take samples and make tests of hypothesis,
c. Repeat parts a and b for a sample size of 25. Did you reject the null hypothesis more often with
a sample size of 10 or a sample size of 25?
■ MINI-PROJECT 9-4
Refer to Case Study 8-1, which discussed the results of a USDA study about the cost of raising a child.
As was noted in that Case Study, the average expenditure to raise a child born in 2008 through the age of
17 for families in each of the three income groups were as shown in the following table.
Family Income
Mean Expenditure
Less than $56,870
$159,870
$56,870 to $98,470
$221,190
More than $98,470
$366,660
a. Recently a random sample of 10 families was selected from the Less than $56,780 group. The
sample mean was found to be $165,000. Determine the value of the sample standard deviation
that would make the p-value of the test H : p = $159,870 versus H { ; jx > $159,870 equal to .03.
What assumption is necessary for your calculations to be valid?
b. Repeat part a for sample sizes of 15, 25, and 50, respectively, using the same sample mean and a
p-value of .03. What can you conclude about the relationship between the sample standard devi-
ation and the sample size in order to maintain the same p-value?
c. Suppose you take a sample of 10 from the income group $56,870 to $98,470, and find the
sample mean to be $226,320. Suppose you are to perform the test: H : p = $221,190 versus
Hi'. /J- > $221,190. Determine the value of the sample standard deviation that would make the
p-value of the test equal to .03. If you were to repeat the procedure for sample sizes of 15, 25,
and 50, respectively, using the same sample mean and a p-value of .03, what do you think your
results would be?
■ MINI-PROJECT 9-5
In Uses and Misuses Part 2 of this Chapter, we discussed how a team of Stanford researchers concluded
that when you flip the coin, whichever side of a coin is facing up when you place it on your thumb to flip
it is more likely to be the side that faces up after the coin is flipped. To test this concept, you are going to
flip a coin 100 times. To simplify the record keeping, you should perform all 100 flips with head facing up
when you place the coin on your thumb to flip it or perform all 100 flips with tail facing up when you place
the coin on your thumb to flip it. Do your best to use the same amount of force each time you flip the coin.
a. If you had head facing up when you placed the coin on your thumb, calculate the sample propor-
tion of flips in which head occurred. (Or you can perform this experiment with tail facing up.)
Perform a test with the null hypothesis that the side that started up will land up 50% of the time
versus the alternative hypothesis that the side that started up will land up more than 50% of the
time. Use a significance level of 5%.
b. The Stanford research group concluded that the side that started up will land up 51% of the time.
Use the data from your 100 flips to test the null hypothesis that the side that started up will land
up 5 1 % of the time versus the alternative hypothesis that the side that started up will not land 5 1 %
of the time. Use a significance level of 5%.
c. Suppose that you were really bored one day and decided to repeat this experiment four times, us-
ing more flips each time, as shown in the following table.
Number of flips in which
the side that started up
Number of flips
also landed up
500
255
1000
510
5000
2550
10000
5100
For each of these four cases, calculate the test statistic and p-value for the hypothesis test described in part
a. Based on your results, what can you conclude about the number of repetitions needed to distinguish be-
tween a result that occurs 50% of the time and the one that occurs 51% of the time?
434 Chapter 9 Hypothesis Tests About the Mean and Proportion
DECIDE FOR YOURSELF
STATISTICAL AND PRACTICAL
SIGNIFICANCE
The hypothesis-testing procedure helps us to make a conclusion
regarding a claim or statement, and often this claim or statement is
about the value of a parameter or the relationship between two or
more parameters. When we reject the null hypothesis, we conclude
that the result is statistically significant at the given significance level
of a. So, what exactly does the term "statistically significant" mean?
Using the single-sample analogy, statistically significant implies that
the value of a point estimator (such as a sample mean or sample pro-
portion) of a parameter is far enough (in terms of the standard devi-
ation or standard error) from the hypothesized value of the parame-
ter so that it falls in the most extreme a X 100% of the area under
the sampling distribution curve.
Now the logical follow-up question is: "What does statistically
significant imply with regard to my specific application?" Unlike the
first question, which has a specific answer, the answer to this question
is: "It depends." In any hypothesis test, one must consider the practi-
cal significance of the result. For example, suppose a new gasoline
additive has been invented and the company that produces it claims
that it increases average gas mileage. A fleet of cars of a specific
model, based on EPA numbers, obtains an average of 448 miles per
tank full of gas without this additive. A random sample of 25 such cars
is selected. Each car is driven on a tank full of gas with this additive
added to the gas. The sample mean for these 25 cars is found to be 453
miles per tank full of gas, with a sample standard deviation of 22
miles. To understand the difference between the statistical significance
and practical significance, find answers to the following questions.
1. Perform the appropriate hypothesis test using the t distribution to
determine if the average mileage per tank full of gas increases with
the additive. Use a 5% significance level. Is this increase statistically
significant? Assume that population is normally distributed.
2. Now suppose we use a sample of 100 cars instead of 25 cars, but
the values of the means and the standard deviation remain the same.
Perform the above hypothesis test again and see if your answer
changes with this larger sample size.
3. Regardless of the sample size, discuss whether the result (453
miles versus 448 miles) is practically significant, that is, whether or
not the increase is meaningful to the everyday driver. Suppose it is
recommended that the additive should be used every 3000 miles.
Assuming that the price of gas is $2.50 per gallon and the gas tank
holds 16 gallons of gas, calculate the savings in gas expenditure per
mile. Then multiply this number by 3000 to obtain the savings per
application of the additive. Assuming that the additive is not free, is
it worth using it?
ECHNOLOGY
INSTRUCTION
Hypothesis Testing
1. To test a hypothesis about a population mean fi given the population standard deviation
cr, select STAT >TESTS >ZTest. If you have the data stored in a list, select Data,
and enter the name of the list. If you have the summary statistics, choose Stats, and
enter the sample mean and size. Enter /j,0, the constant value for the population mean
from your null hypothesis. Enter your value for cr, and select which alternative hypoth-
esis you are using. Select Calculate. (See Screen 9.1.)
2. To test a hypothesis about a population mean p. without knowing the population standard
deviation cr, select STAT >TESTS >TTest. If you have the data stored in a list, select
Data, and enter the name of the list. If you have the summary statistics, choose Stats, and
enter the sample mean, standard deviation, and size. Enter /M, the constant value for the
population mean from your null hypothesis. Select which alternative hypothesis you are
using. Select Calculate.
3. To test a hypothesis about a population proportion p, select STAT >TESTS >l-PropZTest.
Enter the constant value for p from the null hypothesis as pO. Enter the number of successes
as x and the sample size as n. Select the alternative hypothesis you are using. Select
Calculate.
Z-Test
Inpt: Data
u.n:0
<t: 1
x: 3
n:5l
u'.:EHH1 <nn
Calcu 1 ate
Draw
Technology Instruction 435
1. To perform a hypothesis test for the population mean /j, when the population standard devi-
ation o" is given, select Stat >Basic Statistics >1 -Sample Z. If you have your data en-
tered in a column, enter the name of that column in the Samples in columns: box. Instead,
if you know the summary statistics, click next to Summarized data and enter the values
of the Sample size and Mean in their respective boxes. In both cases, enter the value of
the population standard deviation in the Standard deviation box. Enter the value of j±
from the null hypothesis in the Test mean: box (See Screen 9.2). Click on the Options
button and select the appropriate alternative hypothesis from the Alternative box. Click
OK in both windows. The output will appear in the Session window, which will give the
p-value for the test. Based on this ;?-value, you can make a decision.
1 -Sample Z [Test and Confidence Interval)
(* Samples in columns:
CI
(* Summarized data
Sample size: f
Mean:
Standard deviation: |l
Test mean: [tT
(required for test]
Select
Graphs...
Options...
Help
OK
Cancel
2. To perform a hypothesis test for the population mean /jl when the population standard
deviation a is not known, select Stat >Basic Statistics >1-Sample t. If you have your
data entered in a column, enter the name of that column in the Samples in columns:
box. Instead, if you know the summary statistics, click next to Summarized data and
enter the values of the Sample size, Sample standard deviation, and Mean in then-
respective boxes. Enter the value of /j, from the null hypothesis in the Test mean: box.
Click on the Options button, and select the appropriate alternative hypothesis from the
Alternative box. Click OK in both windows. The output will appear in the Session
window, which will give the p-value for the test. Based on this p- value, you can make
a decision.
3. To perform a hypothesis test for the population proportion p, select Stat >Basic
Statistics >1 Proportion. If you have sample data (consisting of values for successes and
failures) entered in a column, enter the name of that column in the Samples in columns:
box. Instead, if you know the number of trials and number of successes, click next to
Summarized data, and enter the required values in the Number of trials: and Number
436 Chapter 9 Hypothesis Tests About the Mean and Proportion
of events: boxes, respectively. Click on the Options button, and enter the value of the pro-
portion from the null hypothesis in the Test proportion: box. Select the appropriate alterna-
tive hypothesis from the Alternative box, and check the box next to Use test and interval
based on normal distribution. Click OK in both windows. The output will appear in the
Session window, which will give the p-value for the test. Based on this p-value, you can
make a decision.
The Data Analysis ToolPak does not contain a preprogrammed function for a test about a
population mean in which the population standard deviation is known. The Excel function
ZTEST works easily only in specific situations, and it requires substantial adjustment in a
number of situations, so it will not be discussed here.
The Data Analysis ToolPak also does not contain a preprogrammed function for a test
about a population mean in which the population standard deviation is unknown. However,
the function used for the paired f-test, which is covered in Chapter 10, can be manipulated
relatively easily in order to produce results for a one-sample r-test. (Note: The Excel function
TTEST has issues similar to the ZTEST function.)
1. Create a second column of data that is the same length as the data that you wish to ana-
lyze. All of the entries in the second column of data should be zero. (See Screen 9.3.)
2. Click the Data tab, then click the Data Analysis button within the Analysis group.
From the Data Analysis window that will appear, select t-test: Paired Two Sample
for Means.
3. Enter the location of the data you wish to analyze in the Variable 1 Range box. Enter the
location of the column of zeroes in the Variable 2 Range box. Enter the value for /jl in the
null hypothesis in the Hypothesized Mean Difference box. Enter the significance level, as
a decimal, in the Alpha box. Choose how you wish the output to appear. (See Screen 9.4.)
Click OK.
A
B
1
data
z
14.3
3
25.2
6]
4
22.5
5
3S.3
ol
6
16.9
7
26.7
o]
8
19.5
9
23.1
10
41
U
33.9
t-Test: Paired Two Sample for Means
Input
Variable 1. Range:
Variable 2 Range:
Hypothesized Mean Difference:
□ Labels
Alpha: 1 0.05
25
Output options
O Output Range:
New Worksheet Ply:
O New Workbook
OK
Cancel
Help
Screen 9.3
Screen 9.4
Technology Assignments 437
A
B
C
1
t-Test: Paired Two Sample for Means
2
3
Variable 1
Variable 2
4
Mean
26.14
5
Variance
80.24933333
6
Observations
10
10
7
Pearson Correlation
#DIV/0S
8
Hypothesized Mean Difference
25
9
df
9
10
tStat
0.402424242
11
P(T<=t) one-tail
0.34S3S114S
12
t Critical one-tail
1.833112923
13
P(T<=t) two-tail
0. S3 6 76229 6
14
t Critical two -tail
2. 26215 715 S
The two lines in the output that you will need
to determine the p-value are the lines labeled t
Stat and P(T<=t) two-tail. (See Screen 9.5.)
If the alternative hypothesis is two-tailed, the
value in the P(T<=t) two-tail box is the p-
value for the test. If the alternative hypothesis
is one-tailed, use the following set of rules:
a. If the hypothesis test is left-tailed and the
value of t Stat is negative OR the
hypothesis test is right-tailed and the value
of t Stat is positive, the ;?-value of the test
is equal to one-half the value in the
P(T<=t) two-tail box.
b. If the hypothesis test is left-tailed and the
value of t Stat is positive OR the
hypothesis test is right-tailed and the value
of t Stat is negative, the p-value of the test
is equal to 1 minus one-half the value in
the P(T<=t) two-tail box.
Screen 9.5
TECHNOLOGY ASSIGNMENTS
TA9.1 According to Freddie Mac, the average rate on a 30-year fixed rate mortgage on June 25, 2009,
was 5.42% (USA TODAY, July, 2, 2009). The following data represent the rates on 50 randomly selected
30-year fixed-rate mortgages granted during the week of July 6 to July 10, 2009.
4.80
5.47
6.36
5.95
5.46
6.18
5.16
5.78
4.67
5.61
5.65
4.97
5.75
5.62
6.17
5.07
5.43
4.65
5.05
5.10
5.06
5.83
5.39
6.09
5.01
5.85
5.10
5.93
5.21
4.80
5.41
4.67
5.58
5.50
5.36
5.54
5.54
5.93
5.84
5.43
5.28
4.74
4.89
5.83
5.86
6.19
4.97
4.73
5.48
5.98
Test at the 5% significance level whether the average interest rate on all 30-year fixed-rate mortgages
granted during the week of July 6 to July 10, 2009 was different from 5.42%.
TA9.2 Some colleges are known for their excellent cafeteria food, so much so that the term "Freshman
15" has been coined to refer to the amount of weight that students gain during their freshman year at col-
lege. The following data represent the amount of weight gained by 45 randomly selected students from a
college during their freshman year. Note that a negative value implies that a student lost weight.
21.1
17.7
25.1
9.8
25.9
5.3
0.3
23.4
22.4
7.6
25.5
15.9
24.2
27.5
-3.0
8.7
13.6
11.2
13.5
7.8
17.8
5.9
-2.4
2.9
0.7
-1.0
25.7
18.0
28.7
3.2
2.2
26.7
24.5
10.5
25.5
-3.2
-0.5
8.0
5.7
-4.6
Although this college is happy about the reputation of its food service, it is concerned about the health is-
sues of substantial weight gains. As a result, it distributed nutrition pamphlets to students in an attempt to
reduce the amount of weight gain. Perform a hypothesis test at the 10% significance level to determine
whether the average weight gain by all freshmen at this college during the first year is less than 15 pounds.
TA9.3 General Logs Banana Bombs cereal is sold in 10.40-ounce packages. Because the cereal is sold
by weight, the number of pieces of Banana Bombs varies from box to box. The following values repre-
sent the number of pieces in 19 boxes of Banana Bombs.
686 695 690 681 683 705 724 701 689 698
715 703 711 676 686 695 697 707 693
438 Chapter 9 Hypothesis Tests About the Mean and Proportion
Perform a hypothesis test to determine whether the average number of pieces in all 10.40-ounce boxes
of Banana Bombs is different from 700. Assume that the distribution of the number of pieces in a
10.40-ounce box is approximately normal. Use a = .05.
TA9.4 According to a basketball coach, the mean height of all male college basketball players is 74 inches.
A random sample of 25 such players produced the following data on their heights.
68 76 74 83 77 76 69 67 71 74 79 85 69
78 75 78 68 72 83 79 82 76 69 70 81
Test at the 2% significance level whether the mean height of all male college basketball players is differ-
ent from 74 inches. Assume that the heights of all male college basketball players are (approximately)
normally distributed.
TA9.5 A past study claimed that adults in America spent an average of 18 hours a week on leisure
activities. A researcher took a sample of 10 adults from a town and asked them about the time they spend
per week on leisure activities. Their responses (in hours) follow.
14 25 22 38 16 26 19 23 41 33
Assume that the times spent on leisure activities by all adults are normally distributed and the population
standard deviation is 3 hours. Using the 5% significance level, can you conclude that the claim of the ear-
lier study is true?
TA9.6 According to a May 14, 2009, Harris Interactive poll (www.harrisinteractive.com), 29% of
Americans drink alcohol at least three times per week. Suppose that in a current survey of 700 Americans,
191 drink alcohol at least three times per week. Perform a hypothesis test to determine whether the per-
centage of Americans who drink alcohol at least three times per week is less than 29%. Use a = .02.
TA9.7 A mail-order company claims that at least 60% of all orders it receives are mailed within 48 hours.
From time to time the quality control department at the company checks if this promise is kept. Recently,
the quality control department at this company took a sample of 400 orders and found that 224 of them
were mailed within 48 hours of the placement of the orders. Test at the 1% significance level whether or
not the company's claim is true.
Chapter
Estimation and Hypothesis Testing:
Two Populations
Are you planning to take a vacation either just to relax, or just to get away from work and school
for a few days? Is taking a vacation very important to you? Vacation may not be equally important
to all of us. A recent survey by Access America suggests that taking vacations is more important
to the younger generation. (See Case Study 10-2).
Chapters 8 and 9 discussed the estimation and hypothesis-testing procedures for /j, and p involving
a single population. This chapter extends the discussion of estimation and hypothesis-testing proce-
dures to the difference between two population means and the difference between two population
proportions. For example, we may want to make a confidence interval for the difference between the
mean prices of houses in California and in New York, or we may want to test the hypothesis that the
mean price of houses in California is different from that in New York. As another example, we may
want to make a confidence interval for the difference between the proportions of all male and female
adults who abstain from drinking, or we may want to test the hypothesis that the proportion of all
adult men who abstain from drinking is different from the proportion of all adult women who abstain
from drinking. Constructing confidence intervals and testing hypotheses about population parameters
are referred to as making inferences.
10.1 Inferences About the
Difference Between Two
Population Means for
Independent Samples:
cr, and <r 2 Known
10.2 Inferences About the
Difference Between
Two Population Means
for Independent
Samples: <r, and <r 2
Unknown but Equal
Case Study 10-1 Average
Compensation for
Accountants
10.3 Inferences About the
Difference Between
Two Population Means
for Independent
Samples: cr, and <r 2
Unknown and Unequal
10.4 Inferences About the
Difference Between
Two Population Means
for Paired Samples
1 0.5 Inferences About the
Difference Between Two
Population Proportions
for Large and
Independent Samples
Case Study 10-2 Is Vacation
Important?
439
440 Chapter 10 Estimation and Hypothesis Testing: Two Populations
10.1 Inferences About the Difference Between
Two Population Means for Independent
Samples: a } and a 2 Known
Let fi l be the mean of the first population and /jl 2 be the mean of the second population. Sup-
pose we want to make a confidence interval and test a hypothesis about the difference between
these two population means, that is, /x, — fi 2 . Let x l be the mean of a sample taken from the
first population and x 2 be the mean of a sample taken from the second population. Then, i, — x 2
is the sample statistic that is used to make an interval estimate and to test a hypothesis about
l±i — jj, 2 . This section discusses how to make confidence intervals and test hypotheses about
fi l — i± 2 when certain conditions (to be explained later in this section) are satisfied. First we
explain the concepts of independent and dependent samples.
10.1.1 Independent Versus Dependent Samples
Two samples are independent if they are drawn from two different populations and the ele-
ments of one sample have no relationship to the elements of the second sample. If the elements
of the two samples are somehow related, then the samples are said to be dependent. Thus, in
two independent samples, the selection of one sample has no effect on the selection of the sec-
ond sample.
Definition
Independent Versus Dependent Samples Two samples drawn from two populations are inde-
pendent if the selection of one sample from one population does not affect the selection of the
second sample from the second population. Otherwise, the samples are dependent.
Examples 10-1 and 10-2 illustrate independent and dependent samples, respectively.
■ EXAMPLE 10-1
Suppose we want to estimate the difference between the mean salaries of all male and all female
executives. To do so, we draw two samples, one from the population of male executives and an-
other from the population of female executives. These two samples are independent because they
are drawn from two different populations, and the samples have no effect on each other. I
■ EXAMPLE 10-2
Suppose we want to estimate the difference between the mean weights of all participants be-
fore and after a weight loss program. To accomplish this, suppose we take a sample of 40 par-
ticipants and measure their weights before and after the completion of this program. Note that
these two samples include the same 40 participants. This is an example of two dependent sam-
ples. Such samples are also called paired or matched samples. I
This section and Sections 10.2, 10.3, and 10.5 discuss how to make confidence intervals
and test hypotheses about the difference between two population parameters when samples are
independent. Section 10.4 discusses how to make confidence intervals and test hypotheses about
the difference between two population means when samples are dependent.
10.1.2 Mean, Standard Deviation, and Sampling
Distribution of x, - x 2
Suppose we select two (independent) samples from two different populations that are referred
to as population 1 and population 2. Let
Illustrating two
independent samples.
Illustrating two
dependent samples.
10.1 Inferences About the Difference Between Two Population Means for Independent Samples: a 1 and <r 2 Known 441
(jl x = the mean of population 1
fi 2 = the mean of population 2
<T[ = the standard deviation of population 1
<x 2 = the standard deviation of population 2
n x = the size of the sample drawn from population 1
n 2 = the size of the sample drawn from population 2
JC] = the mean of the sample drawn from population 1
x 2 = the mean of the sample drawn from population 2
Then, as we discussed in Chapters 8 and 9, if
1. The standard deviation a x of population 1 is known
2. At least one of the following two conditions is fulfilled:
i. The sample is large (i.e., n x > 30)
ii. If the sample size is small, then the population from which the sample is drawn is nor-
mally distributed
then the sampling distribution of x x is normal with its mean equal to /j, x and the standard devi-
ation equal to 0-,/Vnj, assuming that n x IN x s .05.
Similarly, if
1. The standard deviation <x 2 of population 2 is known
2. At least one of the following two conditions is fulfilled:
i. The sample is large (i.e., n 2 > 30)
ii. If the sample size is small, then the population from which the sample is drawn is nor-
mally distributed
then the sampling distribution of x 2 is normal with its mean equal to j± 2 and the standard devi-
ation equal to cr 2 l\/n 2 , assuming that n 2 IN 2 — -05.
Using these results, we can make the following statements about the mean, the standard
deviation, and the shape of the sampling distribution of x x — x 2 .
If the following conditions are satisfied,
1. The two samples are independent
2. The standard deviations cr x and er 2 of the two populations are known
3. At least one of the following two conditions is fulfilled:
i. Both samples are large (i.e., n x > 30 and n 2 > 30)
ii. If either one or both sample sizes are small, then both populations from which the sam-
ples are drawn are normally distributed
then the sampling distribution of x x — x 2 is (approximately) normally distributed with its mean
and standard deviation 1 as, respectively,
In these cases, we can use the normal distribution to make a confidence interval and test a hy-
pothesis about j± l — jx 2 . Figure 10.1 shows the sampling distribution of x x — x 2 when the above
conditions are fulfilled.
'The formula for the standard deviation of x x — x 2 can also be written as
and
where a 7t = aj\fn x and ov = er 2 /V«2.
442 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Sampling Distribution, Mean, and Standard Deviation of x, - x 2 When the conditions listed on
the previous page are satisfied, the sampling distribution of x { — x 2 is (approximately) normal
with its mean and standard deviation as, respectively,
M. . = Mi - M2 and =
Note that to apply the procedures learned in this chapter, the samples selected must be sim-
ple random samples.
10.1.3 Interval Estimation of /a, - fi 2
By constructing a confidence interval for — fi 2 , we find the difference between the means
of two populations. For example, we may want to find the difference between the mean heights
of male and female adults. The difference between the two sample means, x l — x 2 , is the point
estimator of the difference between the two population means, yu, — /jl 2 . When the conditions
mentioned earlier in this section hold true, we use the normal distribution to make a confidence
interval for the difference between the two population means. The following formula gives the
interval estimation for /x, — /jl 2 .
Confidence Interval for /a, — fi 2 When using the normal distribution, the (1 - a) 100% confi-
dence interval for jx x — [i 2 is
(*i - Xl) ± zo-^-i,
The value of z is obtained from the normal distribution table for the given confidence level. The
value of a~x -xi i s calculated as explained earlier. Here, x l — x 2 is the point estimator of fi l — ijl 2 .
Note that in the real world, a t and a 2 are never known. Consequently we will never use
the procedures of this section, but we are discussing these procedures in this book for the in-
formation of the readers.
Example 10-3 illustrates the procedure to construct a confidence interval for — /a 2 us-
ing the normal distribution.
Cor,
ir,
Constructing a confidence
interval for //,, — /i, 2 : (7i
and tr 2 known, and
samples are large.
■ EXAMPLE 10-3
A 2008 survey of low- and middle-income households conducted by Demos, a liberal public
policy group, showed that consumers aged 65 years and older had an average credit card debt
of $10,235 and consumers in the 50- to 64-year age group had an average credit card debt of
$9342 at the time of the survey (USA TODAY, July 28, 2009). Suppose that these averages
were based on random samples of 1200 and 1400 people for the two groups, respectively. Fur-
ther assume that the population standard deviations for the two groups were $2800 and $2500,
respectively. Let and fi 2 be the respective population means for the two groups, people aged
65 years and older and people in the 50- to 64-year age group.
(a) What is the point estimate of fi l — /jl 2 ?
(b) Construct a 97% confidence interval for i± x
fi 2 .
10.1 Inferences About the Difference Between Two Population Means for Independent Samples: a 1 and <r 2 Known 443
Solution Let us refer to consumers aged 65 years and older as population 1 and those in
the 50- to 64-year age group as population 2. Then the respective samples are samples 1
and 2. Let x x and x 2 be the means of the two samples, respectively. From the given infor-
mation,
For 65 and older group: n, = 1200, x, = $10,235, a l = $2800
For 50-64 age group: n 2 = 1400, x 2 = $9342, a 2 = $2500
(a) The point estimate of jx x — /jl 2 is given by the value of x t — x 2 . Thus,
Point estimate of /a, - /x 2 = $10,235 - $9342 = $893
(b) The confidence level is 1 — a = .97. From the normal distribution table, the values
of z for .015 and .9850 areas to the left are —2.17 and 2.17, respectively. Hence, we
will use z = 2.17 in the confidence interval formula. First we calculate the standard
deviation of x l - x 2 , Oj as follows:
[a\ a\ /(2800) 2 (2500) 1
o-j = x — - + — = \ + = $104.8695335
x ' Vn, n 2 V 1200 1400
Next, substituting all the values in the confidence interval formula, we obtain a
97% confidence interval for /j,, — fju 2 as
(xi - x 2 ) ± Z(t- x ^ = ($10,235 - $9342) ± 2.17(104.8695335)
= 893 ± 227.57 = $665.43 to $1120.57
Thus, with 97% confidence we can state that the difference between the means
of 2008 credit card debts for the two groups is between $665.43 and $1120.57. The
value zu^-% = $227.57 is called the margin of error for this estimate. ■
Note that in Example 10-3 both sample sizes were large and the population standard de-
viations were known. If the standard deviations of the two populations are known, at least one
of the sample sizes is small, and both populations are normally distributed, we use the normal
distribution to make a confidence interval for — /jl 2 . The procedure in this case is exactly
the same as in Example 10-3.
10.1.4 Hypothesis Testing About - fi 2
It is often necessary to compare the means of two populations. For example, we may want
to know if the mean price of houses in Chicago is the same as that in Los Angeles. Simi-
larly, we may be interested in knowing if, on average, American children spend fewer hours
in school than Japanese children do. In both these cases, we will perform a test of hypoth-
esis about jii] — fJL 2 . The alternative hypothesis in a test of hypothesis may be that the means
of the two populations are different, or that the mean of the first population is greater than
the mean of the second population, or that the mean of the first population is less than the
mean of the second population. These three situations are described next.
1. Testing an alternative hypothesis that the means of two populations are different is equiv-
alent to # )jl 2 , which is the same as — /jl 2 ¥= 0.
2. Testing an alternative hypothesis that the mean of the first population is greater than the mean
of the second population is equivalent to jx x > jx 2 , which is the same as fi l — i± 2 > 0.
3. Testing an alternative hypothesis that the mean of the first population is less than the mean
of the second population is equivalent to fi l < fi 2 , which is the same as i± x — fi 2 < 0.
The procedure followed to perform a test of hypothesis about the difference between two
population means is similar to the one used to test hypotheses about single-population param-
eters in Chapter 9. The procedure involves the same five steps for the critical- value approach
that were used in Chapter 9 to test hypotheses about jx and p. Here, again, if the following
444 Chapter 10 Estimation and Hypothesis Testing: Two Populations
conditions are satisfied, we will use the normal distribution to make a test of hypothesis about
Mi - ^2-
1. The two samples are independent.
2. The standard deviations a l and cr 2 of the two populations are known.
3. At least one of the following two conditions is fulfilled:
i. Both samples are large (i.e., n l ^ 30 and n 2 ^ 30)
ii. If either one or both sample sizes are small, then both populations from which the sam-
ples are drawn are normally distributed
Test Statistic Z for X, — i 2 When using the normal distribution, the value of the test statistic z
for x x — x 2 is computed as
(xi - x 2 ) - Oj - it 2 )
z =
The value of /jl x — /jl 2 is substituted from H Q . The value of cr^ is calculated as earlier in this section.
Example 10^1 shows how to make a test of hypothesis about (ju i — fju 2 .
■ EXAMPLE 10-4
Refer to Example 10-3 about the average 2008 credit card debts for consumers of two age
groups. Test at the 1% significance level whether the population means for the 2008 credit
card debts for the two groups are different.
Solution From the information given in Example 10—3,
For 65 and older group: n l = 1200, i, = $10,235, o-, = $2800
For 50-64 age group: n 2 = 1400, x 2 = $9342, a 2 = $2500
Let ji, x and /jl 2 be the respective population means for the two groups, people aged 65 years and
older and the ones in the 50- to 64-year age group. Let x x and x 2 be the corresponding sample means.
Step 1. State the null and alternative hypotheses.
We are to test whether the two population means are different. The two possibilities are as follows:
1. The mean 2008 credit card debts for people of the two age groups are not different.
In other words, /jl x = fi 2 , which can be written as jx x — /jl 2 = 0.
2. The mean 2008 credit card debts for people of the two age groups are different. That
is, # j± 2 , which can be written as jju { — /jl 2 J= 0.
Considering these two possibilities, the null and alternative hypotheses are, respectively,
H : i± { — jx 2 = (The two population means are not different.)
H x : fi l — j± 2 (The two population means are different.)
Step 2. Select the distribution to use.
Here, the population standard deviations, o- x and cr 2 , are known, and both samples are large
(«! > 30 and n 2 > 30). Therefore, the sampling distribution of x x — x 2 is approximately nor-
mal, and we use the normal distribution to perform the hypothesis test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is given to be .01. The # sign in the alternative hypothesis indicates
that the test is two-tailed. The area in each tail of the normal distribution curve is a/2 = .01/2
= .005. The critical values of z for .005 and .9950 areas to the left are (approximately) —2.58
and 2.58 from Table IV of Appendix C. These values are shown in Figure 10.2.
Making a two-tailed test
of hypothesis about /ij — jx 2 :
o~, and o~2 are known, and
samples are large.
10.1 Inferences About the Difference Between Two Population Means for Independent Samples: a 1 and <r 2 Known 445
Step 4. Calculate the value of the test statistic.
The value of the test statistic z for x x — x 2 is computed as follows:
aj a 2 2 /(2800) 2 (2500) 1
1 + — = \ I ~ + - = $104.8695335
1200 1400
r From//
(3c, - x 2 ) - U x - fju) (10,235 - 9342) -
z = = = 8.52
104.8695335
Step 5. Make a decision.
Because the value of the test statistic z = 8.52 falls in the rejection region, we reject the
null hypothesis H . Therefore, we conclude that the mean 2008 credit card debts for the two
age groups mentioned in this example are different.
Using the p-Value to Make a Decision
We can use the /9-value approach to make the above decision. To do so, we keep Steps 1 and 2.
Then in Step 3 we calculate the value of the test statistic z (as done in Step 4 above) and find
the p-value for this z from the normal distribution table. In Step 4, the z value for 5c, — x 2 was
calculated to be 8.52. In this example, the test is two-tailed. The ;?-value is equal to twice the
area under the sampling distribution of x x — x 2 to the right of z = 8.52. From the normal dis-
tribution table (Table IV), the area to the right of z = 8.52 is (approximately) zero. Therefore,
the p-value is zero. As we know from Chapter 9, we will reject the null hypothesis for any a
(significance level) that is greater than the /7-value. Consequently, in this example, we will re-
ject the null hypothesis for any a > 0. Since, a = .01 in this example, which is greater than
zero, we reject the null hypothesis. H
EXERCISES
CONCEPTS AND PROCEDURES
10.1 Briefly explain the meaning of independent and dependent samples. Give one example of each.
10.2 Describe the sampling distribution of x x — x 2 for two independent samples when a x and cr 2 are known
and either both sample sizes are large or both populations are normally distributed. What are the mean
and standard deviation of this sampling distribution?
10.3 The following information is obtained from two independent samples selected from two normally
distributed populations.
1?
7.82
2.35
n 2 = 15 x 2 = 5.99 o- 2 = 3.17
446 Chapter 10 Estimation and Hypothesis Testing: Two Populations
a. What is the point estimate of /x, — /x 2 ?
b. Construct a 99% confidence interval for /x, — /x 2 . Find the margin of error for this estimate.
10.4 The following information is obtained from two independent samples selected from two populations.
n, = 650 3c, = 1.05 cr x = 5.22
n 2 = 675 x 2 = 1.54 <r 2 = 6.80
a. What is the point estimate of /x, — /x 2 ?
b. Construct a 95% confidence interval for — /x 2 . Find the margin of error for this estimate.
10.5 Refer to the information given in Exercise 10.3. Test at the 5% significance level if the two popula-
tion means are different.
10.6 Refer to the information given in Exercise 10.4. Test at the 1% significance level if the two popula-
tion means are different.
10.7 Refer to the information given in Exercise 10.4. Test at the 5% significance level if /x, is less than ix 2 .
10.8 Refer to the information given in Exercise 10.3. Test at the 1% significance level if /Xj is greater than /x 2 .
■ APPLICATIONS
10.9 In parts of the eastern United States, whitetail deer are a major nuisance to farmers and homeown-
ers, frequently damaging crops, gardens, and landscaping. A consumer organization arranges a test of two
of the leading deer repellents A and B on the market. Fifty-six unfenced gardens in areas having high con-
centrations of deer are used for the test. Twenty-nine gardens are chosen at random to receive repellent
A, and the other 27 receive repellent B. For each of the 56 gardens, the time elapsed between application
of the repellent and the appearance in the garden of the first deer is recorded. For repellent A, the mean
time is 101 hours. For repellent B, the mean time is 92 hours. Assume that the two populations of elapsed
times have normal distributions with population standard deviations of 15 and 10 hours, respectively.
a. Let tij and /x 2 be the population means of elapsed times for the two repellents, respectively. Find
the point estimate of /x, — /x 2 .
b. Find a 97% confidence interval for /x, — /x 2 .
c. Test at the 2% significance level whether the mean elapsed times for repellents A and B are dif-
ferent. Use both approaches, the critical-value and p-value, to perform this test.
10.10 The U.S. Department of Labor collects data on unemployment insurance payments. Suppose that
during 2009 a random sample of 70 unemployed people in Alabama received an average weekly benefit
of $199.65, whereas a random sample of 65 unemployed people in Mississippi received an average weekly
benefit of $187.93. Assume that the population standard deviations of all weekly unemployment benefits
in Alabama and Mississippi are $32.48 and $26.15, respectively.
a. Let xij and /x 2 be the means of all weekly unemployment benefits in Alabama and Mississippi
paid during 2009, respectively. What is the point estimate of /x, — /x 2 ?
b. Construct a 96% confidence interval for — ix 2 .
c. Using the 4% significance level, can you conclude that the means of all weekly unemployment
benefits in Alabama and Mississippi paid during 2009 are different? Use both approaches to
make this test.
10.11 A local college cafeteria has a self-service soft ice cream machine. The cafeteria provides bowls
that can hold up to 16 ounces of ice cream. The food service manager is interested in comparing the av-
erage amount of ice cream dispensed by male students to the average amount dispensed by female stu-
dents. A measurement device was placed on the ice cream machine to determine the amounts dispensed.
Random samples of 85 male and 78 female students who got ice cream were selected. The sample aver-
ages were 7.23 and 6.49 ounces for the male and female students, respectively. Assume that the popula-
tion standard deviations are 1.22 and 1.17 ounces, respectively.
a. Let and /x 2 be the population means of ice cream amounts dispensed by all male and female
students at this college, respectively. What is the point estimate of — /x 2 ?
b. Construct a 95% confidence interval for /x t — /x 2 .
c. Using the 1% significance level, can you conclude that the average amount of ice cream dis-
pensed by male college students is larger than the average amount dispensed by female college
students? Use both approaches to make this test.
10.12 Employees of a large corporation are concerned about the declining quality of medical services pro-
vided by their group health insurance. A random sample of 100 office visits by employees of this corpo-
ration to primary care physicians during 2004 found that the doctors spent an average of 19 minutes with
10.2 Inferences About the Difference Between Two Population Means for Independent Samples: <r, and <x 2 Unknown but Equal 447
each patient. This year a random sample of 108 such visits showed that doctors spent an average of 15.5
minutes with each patient. Assume that the standard deviations for the two populations are 2.7 and 2.1
minutes, respectively.
a. Construct a 95% confidence interval for the difference between the two population means for
these two years.
b. Using the 2.5% level of significance, can you conclude that the mean time spent by doctors with
each patient is lower for this year than for 2004?
c. What would your decision be in part b if the probability of making a Type I error were zero?
Explain.
10.13 A car magazine is comparing the total repair costs incurred during the first three years on two sports
cars, the T-999 and the XPY. Random samples of 45 T-999s and 51 XPYs are taken. All 96 cars are 3
years old and have similar mileages. The mean of repair costs for the 45 T-999 cars is $3300 for the first
3 years. For the 51 XPY cars, this mean is $3850. Assume that the standard deviations for the two popu-
lations are $800 and $1000, respectively.
a. Construct a 99% confidence interval for the difference between the two population means.
b. Using the 1 % significance level, can you conclude that such mean repair costs are different for
these two types of cars?
c. What would your decision be in part b if the probability of making a Type I error were zero?
Explain.
10.14 The management at New Century Bank claims that the mean waiting time for all customers at its
branches is less than that at the Public Bank, which is its main competitor. A business consulting firm
took a sample of 200 customers from the New Century Bank and found that they waited an average of
4.5 minutes before being served. Another sample of 300 customers taken from the Public Bank showed
that these customers waited an average of 4.75 minutes before being served. Assume that the standard
deviations for the two populations are 1.2 and 1.5 minutes, respectively.
a. Make a 97% confidence interval for the difference between the two population means.
b. Test at the 2.5% significance level whether the claim of the management of the New Century
Bank is true.
c. Calculate the p-value for the test of part b. Based on this p-value, would you reject the null hy-
pothesis if a = .01? What if a = .05?
10.15 Maine Mountain Dairy claims that its 8-ounce low-fat yogurt cups contain, on average, fewer
calories than the 8-ounce low-fat yogurt cups produced by a competitor. A consumer agency wanted
to check this claim. A sample of 27 such yogurt cups produced by this company showed that they con-
tained an average of 141 calories per cup. A sample of 25 such yogurt cups produced by its competi-
tor showed that they contained an average of 144 calories per cup. Assume that the two populations
are normally distributed with population standard deviations of 5.5 and 6.4 calories, repectively.
a. Make a 98% confidence interval for the difference between the mean number of calories in the
8-ounce low-fat yogurt cups produced by the two companies.
b. Test at the 1% significance level whether Maine Mountain Dairy's claim is true.
c. Calculate the p-value for the test of part b. Based on this p-value, would you reject the null hy-
pothesis if a = .005? What if a = .025?
10.2 Inferences About the Difference Between
Two Population Means for Independent
Samples: cr } and cr 2 Unknown but Equal
This section discusses making a confidence interval and testing a hypothesis about the differ-
ence between the means of two populations, /j,, — /j, 2 , assuming that the standard deviations,
crj and <x 2 , of these populations are not known but are assumed to be equal. There are some
other conditions, explained below, that must be fulfilled to use the procedures discussed in this
section.
If the following conditions are satisfied,
1. The two samples are independent
2. The standard deviations <t x and cr 2 of the two populations are unknown, but they can be as-
sumed to be equal, that is, cr x = a 2
448 Chapter 10 Estimation and Hypothesis Testing: Two Populations
3. At least one of the following two conditions is fulfilled:
i. Both samples are large (i.e., n l > 30 and n 2 s 30)
ii. If either one or both sample sizes are small, then both populations from which the sam-
ples are drawn are normally distributed
then we use the t distribution to make a confidence interval and test a hypothesis about the dif-
ference between the means of two populations, — fi 2 .
When the standard deviations of the two populations are equal, we can use cr for both cr x
and cr 2 . Because cr is unknown, we replace it by its point estimator s p , which is called the pooled
sample standard deviation (hence, the subscript p). The value of s p is computed by using the
information from the two samples as follows.
Pooled Standard Deviation for Two Samples The pooled standard deviation for two samples is
computed as
- jjg + (>7 2 - \) S \
Sp V Hi + n 2 - 2
where n l and n 2 are the sizes of the two samples and s\ and si are the variances of the two
samples, respectively. Here s p is an estimator of cr.
In this formula, n x — 1 are the degrees of freedom for sample 1, n 2 — 1 are the degrees of free-
dom for sample 2, and n x + n 2 — 2 are the degrees of freedom for the two samples taken together.
Note that s p is an estimator of the standard deviation, cr, of each of the two populations.
When s p is used as an estimator of cr, the standard deviation crv -^ of X\ — x 2 is estimated
by _ The value of Sj is calculated by using the following formula.
Estimator of the Standard Deviation of X, - X 2 The estimator of the standard deviation of
Now we are ready to discuss the procedures that are used to make confidence intervals and
test hypotheses about /j,, — /jl 2 for small and independent samples selected from two popula-
tions with unknown but equal standard deviations.
10.2.1 Interval Estimation of jii, - /x 2
As was mentioned earlier in this chapter, the difference between the two sample means, x x — x 2 ,
is the point estimator of the difference between the two population means, jx x — pu 2 . The fol-
lowing formula gives the confidence interval for /jl x — pb 2 when the t distribution is used and
the conditions mentioned earlier in this section are fulfilled.
Confidence Interval for /a, - /i 2 The (1 - a) 100% confidence interval for /x, - p, 2 is
(*i - x 2 ) ± ts- x ^ 7xi
where the value of t is obtained from the t distribution table for the given confidence level and
n x + n 2 — 2 degrees of freedom, and Sg -j, i s calculated as explained earlier.
10.2 Inferences About the Difference Between Two Population Means for Independent Samples: tr, and <x 2 Unknown but Equal 449
Example 10-5 describes the procedure to make a confidence interval for — /jl 2 using the
t distribution.
■ EXAMPLE 10-5
A consumer agency wanted to estimate the difference in the mean amounts of caffeine in
two brands of coffee. The agency took a sample of 15 one-pound jars of Brand I coffee
that showed the mean amount of caffeine in these jars to be 80 milligrams per jar with a
standard deviation of 5 milligrams. Another sample of 12 one-pound jars of Brand II cof-
fee gave a mean amount of caffeine equal to 77 milligrams per jar with a standard devia-
tion of 6 milligrams. Construct a 95% confidence interval for the difference between the
mean amounts of caffeine in one-pound jars of these two brands of coffee. Assume that
the two populations are normally distributed and that the standard deviations of the two
populations are equal.
Solution Let /j,, and /x 2 be the mean amounts of caffeine per jar in all 1-pound jars of Brands
I and II, respectively, and let x x and x 2 be the means of the two respective samples. From the
given information,
n x = 15 I] = 80 milligrams s t = 5 milligrams
n 2 = 12 x 2 = 77 milligrams s 2 = 6 milligrams
The confidence level is 1 — a = .95.
Here, cr x and a 2 are unknown but assumed to be equal, the samples are independent
(taken from two different populations), and the sample sizes are small but the two popula-
tions are normally distributed. Hence, we will use the t distribution to make the confidence
interval for fi x — fi 2 as all conditions mentioned in the beginning of this section are sat-
isfied.
First we calculate the standard deviation of x x — x 2 as follows. Note that since it is assumed
that <r l and cr 2 are equal, we will use s p to calculate s^-^.
(n, - l)s\ + [n 2
\)sl
n x + n 2
(15 - 1)(5) 2 + (12 - 1)(6) 2
11/ , / 1 1
— + — = 5.46260011 A /— + —
n, n 2 v ' V 15 12
15 + 12-2
2.11565593
5.46260011
Next, to find the t value from the t distribution table, we need to know the area in each tail
of the t distribution curve and the degrees of freedom:
Area in each tail = a/2 = (1 - .95)/2 = .025
Degrees of freedom = n x + n 2
15 + 12
25
The t value for df = 25 and .025 area in the right tail of the t distribution curve is 2.060. The
95% confidence interval for fi x — i± 2 is
(3c, - x 2 ) ± ts 7ti ^ = (80 - 77) ± 2.060(2.11565593)
= 3 ± 4.36 = -1.36 to 7.36 milligrams
Thus, with 95% confidence we can state that based on these two sample results, the
difference in the mean amounts of caffeine in 1 -pound jars of these two brands of coffee lies
between —1.36 and 7.36 milligrams. Because the lower limit of the interval is negative, it is
possible that the mean amount of caffeine is greater in the second brand than in the first brand
of coffee.
Note that the value of 3c, — x 2 , which is 80 — 77 = 3, gives the point estimate of /jb t — fi 2 .
The value of % _ 5 , which is 4.36, is the margin of error.
Constructing a confidence
interval for fa — /i 2 : two
independent samples, unknown
but equal ctj and ov
450 Chapter 10 Estimation and Hypothesis Testing: Two Populations
10.2.2 Hypothesis Testing About fx x - fx 2
When the conditions mentioned in the beginning of Section 10.2 are satisfied, the t distribution
is applied to make a hypothesis test about the difference between two population means. The
test statistic in this case is t, which is calculated as follows.
Test Statistic t for X, — X 2 The value of the test statistic tfor x x — x 2 is computed as
( = (*i ~ x 2 ) - (/J-, - fl 2 )
The value of jx x — fi 2 in this formula is substituted from the null hypothesis, and % _ s is
calculated as explained earlier in Section 10.2.1.
Examples 10-6 and 10-7 illustrate how a test of hypothesis about the difference between
two population means for small and independent samples that are selected from two popula-
tions with equal standard deviations is conducted using the t distribution.
Making a two-tailed test of
hypothesis about /x. t — fi 2 -'
two independent samples, and
unknown but equal <j\ and <T 2 .
■ EXAMPLE 10-6
A sample of 14 cans of Brand I diet soda gave the mean number of calories of 23 per can with
a standard deviation of 3 calories. Another sample of 16 cans of Brand II diet soda gave the mean
number of calories of 25 per can with a standard deviation of 4 calories. At the 1% significance
level, can you conclude that the mean numbers of calories per can are different for these two
brands of diet soda? Assume that the calories per can of diet soda are normally distributed for
each of the two brands and that the standard deviations for the two populations are equal.
Solution Let jx l and jx 2 be the mean numbers of calories per can for diet soda of Brand I
and Brand II, respectively, and let x, and x 2 be the means of the respective samples. From the
given information,
n, = 14 x, = 23 s, = 3
16
25
The significance level is a
.01.
Step 1. State the null and alternative hypotheses.
We are to test for the difference in the mean numbers of calories per can for the two brands.
The null and alternative hypotheses are, respectively,
H : — j± 2 = (The mean numbers of calories are not different)
H x : — j± 2 ¥= (The mean numbers of calories are different)
Step 2. Select the distribution to use.
Here, the two samples are independent, cr, and cr 2 are unknown but equal, and the sample
sizes are small but both populations are normally distributed. Hence, all conditions mentioned
in the beginning of Section 10.2 are fulfilled. Consequently, we will use the t distribution.
Step 3. Determine the rejection and nonrejection regions.
The # sign in the alternative hypothesis indicates that the test is two-tailed. The signifi-
cance level is .01. Hence,
Area in each tail = a/2 = .01/2 = .005
Degrees of freedom = n x + n 2 — 2 = 14 + 16 — 2 = 28
The critical values of t for df = 28 and .005 area in each tail of the t distribution curve are
—2.763 and 2.763, as shown in Figure 10.3.
10.2 Inferences About the Difference Between Two Population Means for Independent Samples: <x, and <x 2 Unknown but Equal 451
Figure 10.3
-2.763 2.763 '
f f
1 Two critical values of t 1
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for x l — x 2 is computed as follows:
(n, - 1> + ("2 " m (14 - 1)(3) 2 + (16 - 1)(4) 2
— = ,/- — -— = 3.57071421
n, + tu - 2 V 14 + 16-2
11, ,11
— + — = (3.57071421) J— + — = 1.30674760
«i n 2 'M 14 16
From H a
(.v. - x 2 ) - {jm - to) _ (23 - 25) -
fc-i, 1.30674760
1.531
Step 5. Make a decision.
Because the value of the test statistic t = —1.531 for x l — x 2 falls in the nonrejection re-
gion, we fail to reject the null hypothesis. Consequently we conclude that there is no differ-
ence in the mean numbers of calories per can for the two brands of diet soda. The difference
in x l and x 2 observed for the two samples may have occurred due to sampling error only.
Using the p-Value to Make a Decision
We can use the /?-value approach to make the above decision. To do so, we keep Steps 1 and
2 of this example. Then in Step 3, we calculate the value of the test statistic t (as done in Step
4 above) and then find the /?-value for this t from the t distribution table (Table V of Appen-
dix C) or by using technology. In Step 4 above, the f-value for x x — x 2 was calculated to be
— 1.531. In this example, the test is two-tailed. Therefore, the /7-value is equal to twice the
area under the t distribution curve to the left of t = —1.531. If we have access to technology,
we can use it to find the exact p-value, which will be .137. If we use the t distribution table,
we can only find the range for the /?-value. From Table V of Appendix C, for df = 28, the two
values that include 1.531 are 1.313 and 1.701. (Note that we use the positive value of t, al-
though our t is negative.) Thus, the test statistic t = — 1.531 falls between — 1.313 and — 1.701.
The areas in the t distribution table that correspond to 1.313 and 1.701 are .10 and .05, re-
spectively. Because it is a two-tailed test, the /7-value for t = —1.531 is between 2(.10) = .20
and 2(.05) = .10, which can be written as
.10 < />value < .20
As we know from Chapter 9, we will reject the null hypothesis for any a (significance level)
that is greater than the /?-value. Consequently, in this example, we will reject the null hypoth-
esis for any a > .20 using the above range and not reject it for a < .10. If we use technol-
ogy, we will reject the null hypothesis for a > .137. Since a = .01 in this example, which is
smaller than both .10 and .137, we fail to reject the null hypothesis.
■ EXAMPLE 10-7
A sample of 40 children from New York State showed that the mean time they spend watching tel-
evision is 28.50 hours per week with a standard deviation of 4 hours. Another sample of 35 children
452 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Making a right-tailed test of
hypothesis about fi t — /i 2 ;
two independent samples,
o~ } and o~2 unknown but equal,
and both samples are large.
from California showed that the mean time spent by them watching television is 23.25 hours per
week with a standard deviation of 5 hours. Using a 2.5% significance level, can you conclude that
the mean time spent watching television by children in New York State is greater than that for
children in California? Assume that the standard deviations for the two populations are equal.
Solution Let the children from New York State be referred to as population 1 and those
from California as population 2. Let fi x and /j, 2 be the mean time spent watching television by
children in populations 1 and 2, respectively, and let x, and x 2 be the mean time spent watch-
ing television by children in the respective samples. From the given information,
40
35
28.50 hours
23.25 hours
4 hours
5 hours
The significance level is a = .025.
Step 1. State the null and alternative hypotheses.
The two possible decisions are:
2.
The mean time spent watching television by children in New York State is not greater
than that for children in California. This can be written as = /j, 2 or /x, — fi 2 = 0.
The mean time spent watching television by children in New York State is greater than
that for children in California. This can be written as jjb x > ijl 2 or i± x — i± 2 > 0.
Hence, the null and alternative hypotheses are, respectively,
H : jli] - yLt 2 =
H{. j± x - il 2 >
Note that the null hypothesis can also be written as (jl x — ix 2 < 0.
Step 2. Select the distribution to use.
Here, the two samples are independent (taken from two different populations), a x and a 2 are un-
known but assumed to be equal, and both samples are large. Hence, all conditions mentioned in
the beginning of Section 10.2 are fulfilled. Consequently, we use the t distribution to make the test.
Step 3. Determine the rejection and nonrejection regions.
The > sign in the alternative hypothesis indicates that the test is right-tailed. The signifi-
cance level is .025.
Area in the right tail of the t distribution = a = .025
Degrees of freedom = n, + n 2 — 2 = 40 + 35 — 2 = 73
From the t distribution table, the critical value of t for df = 73 and .025 area in the right tail
of the t distribution is 1.993. This value is shown in Figure 10.4.
Figure 10.4
Do not reject H *-
Reject H
p«= .025
1.993
t
Critical value of t
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for x x — x 2 is computed as follows:
/(», - 1)4 + (n 2 - 1)4 1(40 - 1)(4) 2 + (35 - 1)(5) 2
40 + 35
4.49352655
10.2 Inferences About the Difference Between Two Population Means for Independent Samples: <r, and <x 2 Unknown but Equal 453
_s = s„J— + — = (4.49352655) J — + — = 1.04004930
V «i « 2 V 40 35
j- From H a
(jc, - jc 2 ) - (fi! ~ fJh) (28.50 - 23.25) -
t = = = 5.048
5f Sl _^ 1.04004930
Step 5. Make a decision.
Because the value of the test statistic t = 5.048 for x x — x 2 falls in the rejection region (see
Figure 10.4), we reject the null hypothesis H . Hence, we conclude that children in New York
State spend more time, on average, watching TV than children in California.
Using the p-Value to Make a Decision
To use the />value approach to make the above decision, we keep Steps 1 and 2 of this example.
Then in Step 3, we calculate the value of the test statistic t (as done in Step 4 above) and then find
the p-value for this t from the t distribution table (Table V of Appendix C) or by using technology.
In Step 4 above, the f-value for x x — x 2 was calculated to be 5.048. In this example, the test is
right-tailed. Therefore, the /5-value is equal to the area under the t distribution curve to the right of
t = 5.048. If we have access to technology, we can use it to find the exact />value, which will
be .000. If we use the t distribution table, for df = 73, the value of the test statistic t = 5.048 is
larger than 3.206. Therefore, the />value for t = 5.048 is less than .001, which can be written as
p-value < .001
Since we will reject the null hypothesis for any a (significance level) greater than the p-value,
here we reject the null hypothesis because a = .025 is greater than both the p-values, .001
obtained above from the table and .000 obtained by using technology. Note that obtaining the
p-value = .000 from technology does not mean that the p-value is zero. It means that when
it is rounded to three digits after the decimal, it is .000.
Note: What If the Sample Sizes Are Too Large?
In this section, we used the t distribution to make confidence intervals and perform tests of hy-
pothesis about jj, x — i± 2 - When both sample sizes are large, it does not matter how large (over
30) the sample sizes are if we are using technology. However, if we are using the t distribution
table (Table V of Appendix C), this may pose a problem if samples are too large. Table V in
Appendix C goes up to only 75 degrees of freedom. Thus, if the degrees of freedom are larger
than 75, we cannot use Table V to find the critical value(s) of t. As mentioned in Chapters 8
and 9, in such a situation, there are two options:
1. Use the t value from the last row (the row of oo) in Table V.
2. Use the normal distribution as an approximation to the t distribution.
A few of the exercises at the end of this section present such situations.
EXERCISES
CONCEPTS AND PROCEDURES
10.16 Explain what conditions must hold true to use the t distribution to make a confidence interval and
to test a hypothesis about /jl x — /j, 2 for two independent samples selected from two populations with un-
known but equal standard deviations.
10.17 The following information was obtained from two independent samples selected from two normally
distributed populations with unknown but equal standard deviations.
21
13.97
3.78
n 2 = 20 x 2 = 15.55 s 2 = 3.26
a. What is the point estimate of ix x — /j, 2 ? b. Construct a 95% confidence interval for — n 2 .
AVERAGE
COMPENSATION
FOR
ACCOUNTANTS
USA TODAY Snapshots®
Accountant certificates
and compensation
1 2008 average compensations
Certified management accountant *LS , faOD
Certified public accountant
Both CMAand CPA *150^
Source- inaniiKOIMJnWnwi" \
Atcouimms samysww
By Jae Yang Jnd Julie Sn i der. USA TODAY
The above chart shows the average compensations for the year 2008 for accountants with and without cer-
tification. For example, accountants with no certificate earned an average of $95,974 in 2008, certified man-
agement accountants earned an average of $125,600, and so on. These results are based on an Institute
of Management Accountants survey. If we know the sample sizes and standard deviations for the four cat-
egories listed in the chart, we can make a confidence interval and perform a test of hypothesis for the dif-
ference in the mean compensations for any two groups out of these four groups using the procedures
learned in this section.
Consider the two groups of accountants named certified management accountants (CMA) and certi-
fied public accountants (CPA). Let />, and /j, 2 be the mean compensations for the CMA and CPA groups,
respectively. Let x, and x 2 be the corresponding sample means. Suppose the sample standard deviations
of 2008 compensations for the two groups were $16,000 and $18,000, respectively. Also, assume that
although the population standard deviations are not known but they are (approximately) equal. Suppose that
the sample sizes for these two groups were 900 and 1000, respectively. Then, from the given information:
For CMA: n, = 900 x, = $125,600 s, = $16,000
ForCPA: n 2 = 1000 x 2 = $135,101 s 2 = $18,000
Below we make a confidence interval for and test a hypothesis about /x, - /j. 2 for the CMA and CPA groups.
1. Confidence interval for fi t - fi 2
Suppose we want to make a 98% confidence interval for ^, - ^ 2 . The area in each tail of the r distribu-
tion and the degrees of freedom are
Area in each tail = a/2 = (1 - .98)/2 = .01
Degrees of freedom = n, + n 2 - 2 = 900 + 1000 - 2 = 1898
Because df = 1 898 is not in the f distribution table, we will use the last row of Table V to obtain the t value
for .01 area in the right tail. This f value is 2.326.
10.18 The following information was obtained from two independent samples selected from two popula-
tions with unknown but equal standard deviations.
n x = 55 Xi = 90.40 s, = 11.60
n 2 = 50 x 2 = 86.30 s 2 = 10.25
a. What is the point estimate of — jjL 2 l
b. Construct a 99% confidence interval for — ix 2 .
454
We calculate the standard deviation of x, - x, as follows.
(n, - l)s? + (n 2 - l)sl
(900 - 1)(16,000) 2 + (1000 - 1)(18,000) 2
900 + 1 000 - 2
$17,081.9015
1
S p J— + — = (17,081.9015)
1
900
1
1000
$784.8592
Hence, the 98% confidence interval for /x, - ii 2 is
(/i, - ti 2 ) ± f s,,.^ = (125,600 - 135,101) ± 2.326(784.8592) = -9501 ± 1825.58
= -$11,326.58 to -$7675.42
Thus, the 98% confidence interval for ii, - ii 2 is -$11,326.58 to -$7675.42. In other words, the
average compensation of the CMA group is anywhere between $7675.42 and $11,326.58 less than the
average compensation of the CPA group at a 98% confidence level.
2. Test of hypothesis test about /x, — /a 2
Suppose we want to test, at the 1% significance level, whether the 2008 mean compensation for the CMA
group is less than that for the CPA group. In other words, we are to test if /j,, is less than ii 2 . The null and
alternative hypotheses are
H Q : [J., = fi 2 or /a, - /i 2 =
Hi, Hi < i±2 or Mi — M2 ^
Note that the test is left-tailed. Because the population standard deviations are not known, we will use the
t distribution. The area in the left tail of the t distribution and the degrees of freedom are
Area in the left tail = a = .01
Degrees of freedom = n, + n ;
900 + 1000
1898
Because df = 1898 is not in the f distribution table, we will use the last row of Table V to obtain the f value
for .01 area in the left tail. This f value is -2.326.
As calculated above, the standard deviation of x, - x 2 is
= $784.8592
The value of the test statistic f for x, - x 2 is computed as follows.
(x, -x 2 ) - (/a, - /jl 2 ) (125,600 - 135,101) -
S~ 784.8592
From H„
-12.105
Because the value of the test statistic t = -12.105 is smaller than the critical value of f = -2.326, it falls
in the rejection region. Consequently, we reject the null hypothesis and conclude that the 2008 mean com-
pensation for the CMA group is less than that for the CPA group.
We can also use the p-value approach to make this decision. In this example, the test is left-tailed. As
calculated above, the f value for x, - x 2 is -12.105. From the last row of the f distribution table, -12.105
is less than -3.090. Therefore, the p-value is less than .001. Since, a = .01 in this example is greater than
this p-value of .001, we reject the null hypothesis and conclude that the 2008 mean compensation for the
CMA group is less than that for the CPA group.
Source: The chart reproduced with
permission from USA TODAY, August
19, 2009. Copyright © 2009, USA
TODAY.
10.19 Refer to the information given in Exercise 10.17. Test at the 5% significance level if the two pop-
ulation means are different.
10.20 Refer to the information given in Exercise 10.18. Test at the 1% significance level if the two pop-
ulation means are different.
10.21 Refer to the information given in Exercise 10.17. Test at the 1% significance level if is less than p, 2 -
10.22 Refer to the information given in Exercise 10.18. Test at the 5% significance level if is greater
than /x 2 .
455
456 Chapter 10 Estimation and Hypothesis Testing: Two Populations
10.23 The following information was obtained from two independent samples selected from two normally
distributed populations with unknown but equal standard deviations.
Sample 1: 47.7 46.9 51.9 34.1 65.8 61.5 50.2 40.8 53.1 46.1 47.9 45.7 49.0
Sample 2: 50.0 47.4 32.7 48.8 54.0 46.3 42.5 40.8 39.0 68.2 48.5 41.8
a. Let fi x be the mean of population 1 and /jl 2 be the mean of population 2. What is the point estimate
of fli - /x 2 ?
b. Construct a 98% confidence interval for — ix 2 .
c. Test at the 1% significance level if is greater than /jl 2 .
10.24 The following information was obtained from two independent samples selected from two normally
distributed populations with unknown but equal standard deviations.
Sample 1: 2.18 2.23 1.96 2.24 2.72 1.87 2.68 2.15 2.49 2.05
Sample 2: 1.82 1.26 2.00 1.89 1.73 2.03 1.43 2.05 1.54 2.50 1.99 2.13
a. Let /i, be the mean of population 1 and /jl 2 be the mean of population 2. What is the point estimate
of - /J, 2 ?
b. Construct a 99% confidence interval for — /j, 2 .
e. Test at the 2.5% significance level if jju x is lower than /jl 2 .
■ APPLICATIONS
10.25 The standard recommendation for automobile oil changes is once every 3000 miles. A local mechanic
is interested in determining whether people who drive more expensive cars are more likely to follow the rec-
ommendation. Independent random samples of 45 customers who drive luxury cars and 40 customers who
drive compact lower-price cars were selected. The average distance driven between oil changes was 3187 miles
for the luxury car owners and 3214 miles for the compact lower-price cars. The sample standard deviations
were 42.40 and 50.70 miles for the luxury and compact groups, respectively. Assume that the population dis-
tributions of the distances between oil changes have the same standard deviation for the two populations.
a. Construct a 95% confidence interval for the difference in the mean distances between oil changes
for all luxury cars and all compact lower-price cars.
b. Using the 1% significance level, can you conclude that the mean distance between oil changes
is less for all luxury cars than that for all compact lower-price cars?
10.26 A town that recently started a single-stream recycling program provided 60-gallon recycling bins to 25
randomly selected households and 75-gallon recycling bins to 22 randomly selected households. The total vol-
ume of recycling over a 10-week period was measured for each of the households. The average total volumes
were 382 and 415 gallons for the households with the 60- and 75-gallon bins, respectively. The sample stan-
dard deviations were 52.5 and 43.8 gallons, respectively. Assume that the 10-week total volumes of recycling
are approximately normally distributed for both groups and that the population standard deviations are equal.
a. Construct a 98% confidence interval for the difference in the mean volumes of 10-week recy-
cling for the households with the 60- and 75-gallon bins.
b. Using the 2% significance level, can you conclude that the average 10- week recycling volume
of all households having 60-gallon containers is different from the average volume of all house-
holds that have 75-gallon containers?
10.27 An insurance company wants to know if the average speed at which men drive cars is greater than
that of women drivers. The company took a random sample of 27 cars driven by men on a highway and
found the mean speed to be 72 miles per hour with a standard deviation of 2.2 miles per hour. Another
sample of 1 8 cars driven by women on the same highway gave a mean speed of 68 miles per hour with
a standard deviation of 2.5 miles per hour. Assume that the speeds at which all men and all women drive
cars on this highway are both normally distributed with the same population standard deviation.
a. Construct a 98% confidence interval for the difference between the mean speeds of cars driven
by all men and all women on this highway.
b. Test at the 1% significance level whether the mean speed of cars driven by all men drivers on
this highway is greater than that of cars driven by all women drivers.
10.28 A high school counselor wanted to know if tenth-graders at her high school tend to have more free
time than the twelfth-graders. She took random samples of 25 tenth-graders and 23 twelfth-graders. Each
student was asked to record the amount of free time he or she had in a typical week. The mean for the tenth-
graders was found to be 29 hours of free time per week with a standard deviation of 7.0 hours. For the twelfth-
graders, the mean was 22 hours of free time per week with a standard deviation of 6.2 hours. Assume that
the two populations are normally distributed with equal but unknown population standard deviations.
a. Make a 90% confidence interval for the difference between the corresponding population means.
b. Test at the 5% significance level whether the two population means are different.
10.3 Inferences About the Difference Between Two Population Means for Independent Samples: cr, and <r 2 Unknown and Unequal 457
10.29 A company claims that its medicine, Brand A, provides faster relief from pain than another
company's medicine, Brand B. A researcher tested both brands of medicine on two groups of randomly
selected patients. The results of the test are given in the following table. The mean and standard deviation
of relief times are in minutes.
Mean of
Standard Deviation
Brand
Sample Size
Relief Times
of Relief Times
A
25
44
11
B
22
49
9
a. Construct a 99% confidence interval for the difference between the mean relief times for the two
brands of medicine.
b. Test at the 1 % significance level whether the mean relief time for Brand A is less than that for
Brand B.
Assume that the two populations are normally distributed with unknown but equal standard deviations.
10.30 A consumer organization tested two paper shredders, the Piranha and the Crocodile, designed for home
use. Each of 10 randomly selected volunteers shredded 100 sheets of paper with the Piranha, and then an-
other sample of 10 randomly selected volunteers each shredded 100 sheets with the Crocodile. The Piranha
took an average of 203 seconds to shred 100 sheets with a standard deviation of 6 seconds. The Crocodile
took an average of 187 seconds to shred 100 sheets with a standard deviation of 5 seconds. Assume that the
shredding times for both machines are normally distributed with equal but unknown standard deviations.
a. Construct a 99% confidence interval for the difference between the two population means.
b. Using the 1% significance level, can you conclude that the mean time taken by the Piranha to
shred 100 sheets is greater than that for the Crocodile?
c. What would your decision be in part b if the probability of making a Type I error were zero?
Explain.
10.31 Quadro Corporation has two supermarket stores in a city. The company's quality control depart-
ment wanted to check if the customers are equally satisfied with the service provided at these two stores.
A sample of 380 customers selected from Supermarket I produced a mean satisfaction index of 7.6 (on a
scale of 1 to 10, 1 being the lowest and 10 being the highest) with a standard deviation of .75. Another
sample of 370 customers selected from Supermarket II produced a mean satisfaction index of 8.1 with a
standard deviation of .59. Assume that the customer satisfaction index for each supermarket has unknown
but same population standard deviation.
a. Construct a 98% confidence interval for the difference between the mean satisfaction indexes
for all customers for the two supermarkets.
b. Test at the 1 % significance level whether the mean satisfaction indexes for all customers for the
two supermarkets are different.
10.32 According to the credit rating firm Equifax, credit limits on newly issued credit cards decreased be-
tween 2008 and the period of January to April 2009 {USA TODAY, July 7, 2009). Suppose that random
samples of 200 credit cards issued in 2008 and 200 credit cards issued during the first 4 months of 2009
had average credit limits of $4710 and $4602, respectively, which are comparable to the values given in
the article. Although no information about standard deviations was provided, suppose that the sample stan-
dard deviations for the 2008 and 2009 samples were $485 and $447, respectively, and that the assumption
that the population standard deviations are equal for the two time periods is reasonable.
a. Construct a 95% confidence interval for the difference in the mean credit limits for all new credit
cards issued in 2008 and during the first 4 months of 2009.
b. Using the 2.5% significance level, can you conclude that the average credit limit for all new credit
cards issued in 2008 was higher than the corresponding average for the first 4 months of 2009?
10.3 Inferences About the Difference Between
Two Population Means for Independent
Samples: a } and a 2 Unknown and Unequal
Section 10.2 explained how to make inferences about the difference between two population
means using the t distribution when the standard deviations of the two populations are unknown
but equal and certain other assumptions hold true. Now, what if all other assumptions of Sec-
tion 10.2 hold true, but the population standard deviations are not only unknown but also
458 Chapter 10 Estimation and Hypothesis Testing: Two Populations
unequal? In this case, the procedures used to make confidence intervals and to test hypotheses
about i± x — jx 2 remain similar to the ones we learned in Sections 10.2.1 and 10.2.2, except for
two differences. When the population standard deviations are unknown and not equal, the de-
grees of freedom are no longer given by n, + n 2 — 2, and the standard deviation of x x — x 2 is
not calculated using the pooled standard deviation s p .
Degrees of Freedom If
1. The two samples are independent
2. The standard deviations a { and <j 2 of the two populations are unknown and unequal, that
is, o - ! cr 2
3. At least one of the following two conditions is fulfilled:
i. Both samples are large (i.e., n l > 30 and n 2 s 30)
ii. If either one or both sample sizes are small, then both populations from which the samples
are drawn are normally distributed
then the f distribution is used to make inferences about fi l — /jl 2 , and the degrees of freedom for
the t distribution are given by
The number given by this formula is always rounded down for df.
Because the standard deviations of the two populations are not known, we use Sj as a
point estimator of Oj The following formula is used to calculate the standard deviation
of X] — x 2 .
Estimate of the Standard Deviation of x, - x 2 The value of s Si _^ is calculated as
10.3.1 Interval Estimation of /a, - (jl 2
Again, the difference between the two sample means, x } — x 2 , is the point estimator of the dif-
ference between the two population means, — /jl 2 . The following formula gives the confi-
dence interval for /a, — /j, 2 when the t distribution is used and the conditions mentioned earlier
in this section are satisfied.
Confidence Interval for /i, - /i 2 The (1 - a) 100% confidence interval for fii - /jl 2 is
(X[ — X 2 ) ± fS^-j,
where the value of t is obtained from the t distribution table for a given confidence level and
the degrees of freedom are given by the formula mentioned earlier, and _= is also calculated
as explained earlier.
Example 10-8 describes how to construct a confidence interval for jx x — /jl 2 when the stan-
dard deviations of the two populations are unknown and unequal.
10.3 Inferences About the Difference Between Two Population Means for Independent Samples: cr, and <r 2 Unknown and Unequal 459
■ EXAMPLE 10-8
According to Example 10-5 of Section 10.2.1, a sample of 15 one-pound jars of coffee of
Brand I showed that the mean amount of caffeine in these jars is 80 milligrams per jar with a
standard deviation of 5 milligrams. Another sample of 12 one-pound coffee jars of Brand II
gave a mean amount of caffeine equal to 77 milligrams per jar with a standard deviation of
6 milligrams. Construct a 95% confidence interval for the difference between the mean amounts
of caffeine in one-pound coffee jars of these two brands. Assume that the two populations are
normally distributed and that the standard deviations of the two populations are not equal.
Constructing a confidence
interval for /u,[ — fi 2 -' two
independent samples, (7i and
tr 2 unknown and unequal.
Solution Let and /jl 2 be the mean amounts of caffeine per jar in all 1-pound jars of Brands
I and II, respectively, and let x l and x 2 be the means of the two respective samples.
From the given information,
n ] = 15 3c ! = 80 milligrams Si = 5 milligrams
n 2 = 12 3c 2 = 77 milligrams s 2 = 6 milligrams
The confidence level is 1 — a = .95.
First, we calculate the standard deviation of 3c i — x 2 as follows:
2 2
ft +
«j n 2
(5f
15
+
(6f
12
2.16024690
Next, to find the t value from the t distribution table, we need to know the area in each tail
of the t distribution curve and the degrees of freedom:
Area in each tail = a/2 = (1 - .95)/2 = .025
15 12
df-
4
2\2
s 2 ^
2\2
*2 N
+
1 n 2
1
(5) 2 V
15
15 -
+
1 12
(6fy
12 )_
1
21.42 « 21
Note that the degrees of freedom are always rounded down as in this calculation. From the t
distribution table, the t value for df = 21 and .025 area in the right tail of the t distribution
curve is 2.080. The 95% confidence interval for /j,, — jx 2 is
(*, - x 2 )
(80 - 77)
3 ± 4.49 :
: 2.080(2.16024690)
-1.49 to 7.49
Thus, with 95% confidence we can state that based on these two sample results, the differ-
ence in the mean amounts of caffeine in 1 -pound jars of these two brands of coffee is between
— 1.49 and 7.49 milligrams. H
Comparing this confidence interval with the one obtained in Example 10-5, we observe
that the two confidence intervals are very close. From this we can conclude that even if the
standard deviations of the two populations are not equal and we use the procedure of Section
10.2.1 to make a confidence interval for /a, — jx 2 , the margin of error will be small as long
as the difference between the two population standard deviations is not too large.
10.3.2 Hypothesis Testing About /a, - fx 2
When the standard deviations of the two populations are unknown and unequal along with the other
conditions of Section 10.3 holding true, we use the t distribution to make a test of hypothesis about
l± x — ji, 2 . This procedure differs from the one in Section 10.2.2 only in the calculation of degrees
of freedom for the t distribution and the standard deviation of 3cj — x 2 . The df and the standard de-
viation of 3c] — x 2 in this case are given by the formulas used in Section 10.3.1.
460 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Test Statistic t for X, — X 2 The value of the test statistic tforx l — x 2 is computed as
(*i - x 2 ) - - fl 2 )
t =
The value of fi l — /jl 2 in this formula is substituted from the null hypothesis, and s^_^ is
calculated as explained earlier.
Example 10-9 illustrates the procedure used to conduct a test of hypothesis about
— jx 2 when the standard deviations of the two populations are unknown and unequal.
Making a two-tailed test of
hypothesis about /ii — fi 2 -
two independent samples,
and unknown and unequal
o- l and £r 2 .
■ EXAMPLE 10-9
According to Example 10-6 of Section 10.2.2, a sample of 14 cans of Brand I diet soda gave
the mean number of calories per can of 23 with a standard deviation of 3 calories. Another
sample of 16 cans of Brand II diet soda gave the mean number of calories of 25 per can with
a standard deviation of 4 calories. Test at the 1% significance level whether the mean num-
bers of calories per can of diet soda are different for these two brands. Assume that the calo-
ries per can of diet soda are normally distributed for each of these two brands and that the
standard deviations for the two populations are not equal.
Solution Let ix x and /x 2 be the mean numbers of calories for all cans of diet soda of Brand
I and Brand II, respectively, and let jc, and x 2 be the means of the respective samples. From
the given information,
14
16
23
25
The significance level is a
.01.
3
4
Step 1. State the null and alternative hypotheses.
We are to test for the difference in the mean numbers of calories per can for the two brands.
The null and alternative hypotheses are, respectively,
H : fjLx — i± 2 = (The mean numbers of calories are not different.)
Hi. /jii — fi 2 =h (The mean numbers of calories are different.)
Step 2. Select the distribution to use.
Here, the two samples are independent, a l and cr 2 are unknown and unequal, the sample
sizes are small, but both populations are normally distributed. Hence, all conditions mentioned
in the beginning of Section 10.3 are fulfilled. Consequently, we use the t distribution to make
the test.
Step 3. Determine the rejection and nonrejection regions.
The ¥= sign in the alternative hypothesis indicates that the test is two-tailed. The signifi-
cance level is .01. Hence,
Area in each tail = a/2 = .01/2 = .005
The degrees of freedom are calculated as follows:
df-
2 2\?
"i n 2/
(3f
14
(4fy
16 J
t 2\2
*1
t 2\2
(3f\2
14
14 - 1
+
(4^ 2
16
27.41 « 27
16
1
10.3 Inferences About the Difference Between Two Population Means for Independent Samples: cr, and d 2 Unknown and Unequal 461
From the t distribution table, the critical values of t for df = 27 and .005 area in each tail of
the t distribution curve are —2.771 and 2.771. These values are shown in Figure 10.5.
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for x x — x 2 is computed as follows:
TTi /(3) 2 . (4) 2
./— + — = 1.28173989
j- From H
(xi ~ x 2 ) ~ Oi - f* 2 ) (23 - 25) -
t = = = -1.560
s Sl _^ 1.28173989
Step 5. Make a decision.
Because the value of the test statistic t = —1.560 for x x — x 2 falls in the nonrejection re-
gion, we fail to reject the null hypothesis. Hence, there is no difference in the mean numbers
of calories per can for the two brands of diet soda. The difference in x x and x 2 observed for
the two samples may have occurred due to sampling error only.
Using the p-Value to Make a Decision
We can use the p-value approach to make the above decision. To do so, we keep Steps 1 and 2
of this example. Then in Step 3 we calculate the value of the test statistic t (as done in Step 4
above) and then find the p- value for this t from the t distribution table (Table V of Appendix C)
or by using technology. In Step 4 above, the t- value for x x — x 2 was calculated to be — 1.560. In
this example, the test is two-tailed. Therefore, the p-value is equal to twice the area under the t
distribution curve to the left of t = — 1 .560. If we have access to technology, we can use it to find
the exact p-value, which will be .130. If we use the t distribution table, we can only find the range
for the p-value. From Table V of Appendix C, for df = 27, the two values that include 1.560 are
1.314 and 1.703. (Note that we use the positive value of t, although our t is negative.) Thus, test
statistic t = —1.560 falls between —1.314 and —1.703. The areas in the t distribution table that
correspond to 1.314 and 1.703 are .10 and .05, respectively. Because it is a two-tailed test, the
p-value for t = —1.560 is between 2(.10) = .20 and 2(.05) = .10, which can be written as
.10 < p-value < .20
Since we will reject the null hypothesis for any a (significance level) that is greater than the
p-value, we will reject the null hypothesis in this example for any a > .20 using the above
range and not reject for a < .10. If we use technology, we will reject the null hypothesis for
a > .130. Since a = .01 in this example, which is smaller than both .10 and .130, we fail to
reject the null hypothesis. ■
The degrees of freedom for the procedures to make a confidence interval and to test a hypoth- -4 Remember
esis about p t — p, 2 learned in Sections 10.3.1 and 10.3.2 are always rounded down.
462 Chapter 10 Estimation and Hypothesis Testing: Two Populations
EXERCISES
CONCEPTS AND PROCEDURES
10.33 Assuming that the two populations are normally distributed with unequal and unknown population
standard deviations, construct a 95% confidence interval for /j,, — /j, 2 for the following.
14 x, = 109.43 s,
2.26
« 2 = 15 x 2 = 113.:
s 7 = 5.84
10.34 Assuming that the two populations have unequal and unknown population standard deviations, con-
struct a 99% confidence interval for — /jl 2 f° r me following.
n x = 48 3c, = .863 Sj = .176
n 2 = 46 x 2 = .796 s 2 = .068
10.35 Refer to Exercise 10.33. Test at the 5% significance level if the two population means are different.
10.36 Refer to Exercise 10.34. Test at the 1% significance level if the two population means are different.
10.37 Refer to Exercise 10.33. Test at the 1% significance level if ix { is less than /j, 2 -
10.38 Refer to Exercise 10.34. Test at the 2.5% significance level if is greater than /j, 2 -
■ APPLICATIONS
10.39 According to the information given in Exercise 10.25, a sample of 45 customers who drive luxury
cars showed that their average distance driven between oil changes was 3187 miles with a sample stan-
dard deviation of 42.40 miles. Another sample of 40 customers who drive compact lower-price cars re-
sulted in an average distance of 3214 miles with a standard deviation of 50.70 miles. Suppose that the
standard deviations for the two populations are not equal.
a. Construct a 95% confidence interval for the difference in the mean distance between oil changes
for all luxury cars and all compact lower- price cars.
b. Using the 1% significance level, can you conclude that the mean distance between oil changes
is lower for all luxury cars than for all compact lower-price cars?
c. Suppose that the sample standard deviations were 28.9 and 61.4 miles, respectively. Redo parts
a and b. Discuss any changes in the results.
10.40 As mentioned in Exercise 10.26, a town that recently started a single-stream recycling program pro-
vided 60-gallon recycling bins to 25 randomly selected households and 75-gallon recycling bins to 22 ran-
domly selected households. The average total volumes of recycling over a 10-week period were 382 and
415 gallons for the two groups, respectively, with standard deviations of 52.5 and 43.8 gallons, respec-
tively. Suppose that the standard deviations for the two populations are not equal.
a. Construct a 98% confidence interval for the difference in the mean volumes of 10-week recy-
cling for the households with the 60- and 75-gallon bins.
b. Using the 2% significance level, can you conclude that the average 10- week recycling volume
of all households having 60-gallon containers is different from the average 10- week recycling
volume of all households that have 75-gallon containers?
c. Suppose that the sample standard deviations were 59.3 and 33.8 gallons, respectively. Redo parts
a and b. Discuss any changes in the results.
10.41 According to Exercise 10.27, an insurance company wants to know if the average speed at which
men drive cars is higher than that of women drivers. The company took a random sample of 27 cars driven
by men on a highway and found the mean speed to be 72 miles per hour with a standard deviation of
2.2 miles per hour. Another sample of 18 cars driven by women on the same highway gave a mean speed
of 68 miles per hour with a standard deviation of 2.5 miles per hour. Assume that the speeds at which all
men and all women drive cars on this highway are both normally distributed with unequal population stan-
dard deviations.
a. Construct a 98% confidence interval for the difference between the mean speeds of cars driven
by all men and all women on this highway.
b. Test at the 1% significance level whether the mean speed of cars driven by all men drivers on
this highway is higher than that of cars driven by all women drivers.
c. Suppose that the sample standard deviations were 1.9 and 3.4 miles per hour, respectively. Redo
parts a and b. Discuss any changes in the results.
10.3 Inferences About the Difference Between Two Population Means for Independent Samples: cr, and d 2 Unknown and Unequal 463
10.42 Refer to Exercise 10.28. Now assume that the two populations are normally distributed with un-
equal and unknown population standard deviations.
a. Make a 90% confidence interval for the difference between the corresponding population
means.
b. Test at the 5% significance level whether the two population means are different.
c. Suppose that the sample standard deviations were 9.5 and 5.1 hours, respectively. Redo parts a
and b. Discuss any changes in the results.
10.43 As mentioned in Exercise 10.29, a company claims that its medicine. Brand A, provides faster re-
lief from pain than another company's medicine, Brand B. A researcher tested both brands of medicine
on two groups of randomly selected patients. The results of the test are given in the following table. The
mean and standard deviation of relief times are in minutes.
Mean of
Standard Deviation
Brand
Sample Size
Relief Times
of Relief Times
A
25
44
11
B
22
49
9
a. Construct a 99% confidence interval for the difference between the mean relief times for the two
brands of medicine.
b. Test at the 1% significance level whether the mean relief time for Brand A is less than that for
Brand B.
c. Suppose that the sample standard deviations were 13.3 and 7.2 minutes, respectively. Redo parts
a and b. Discuss any changes in the results.
Assume that the two populations are normally distributed with unknown and unequal standard deviations.
10.44 Refer to Exercise 10.30. Now assume that the shredding times for both paper shredders are nor-
mally distributed with unequal and unknown standard deviations.
a. Construct a 99% confidence interval for the difference between the two population means.
b. Using the 1% significance level, can you conclude that the mean time taken by the Piranha to
shred 100 sheets is greater than that for the Crocodile?
c. Suppose that the sample standard deviations were 7.40 and 4.60 seconds, respectively. Redo parts
a and b. Discuss any changes in the results.
d. What would your decision be in part b if the probability of making a Type I error were zero?
Explain.
10.45 As mentioned in Exercise 10.31, Quadro Corporation has two supermarkets in a city. The com-
pany's quality control department wanted to check if the customers are equally satisfied with the service
provided at these two stores. A sample of 380 customers selected from Supermarket I produced a mean
satisfaction index of 7.6 (on a scale of 1 to 10, 1 being the lowest and 10 being the highest) with a stan-
dard deviation of .75. Another sample of 370 customers selected from Supermarket II produced a mean
satisfaction index of 8.1 with a standard deviation of .59. Assume that the customer satisfaction index for
each supermarket has an unknown and different population standard deviation.
a. Construct a 98% confidence interval for the difference between the mean satisfaction indexes
for all customers for the two supermarkets.
b. Test at the 1% significance level whether the mean satisfaction indexes for all customers for the
two supermarkets are different.
c. Suppose that the sample standard deviations were .88 and .39, respectively. Redo parts a and b.
Discuss any changes in the results.
10.46 Refer to Exercise 10.32. It was mentioned that the average credit limit on 200 credit cards issued
during 2008 was $4710 with a standard deviation of $485, and the mean and standard deviation for 200
credit cards issued during the first 4 months of 2009 were $4602 and $447, respectively. Suppose that the
standard deviations for the two populations are not equal.
a. Construct a 95% confidence interval for the difference in the mean credit limits for all new credit
cards issued in 2008 and during the first 4 months of 2009.
b. Using the 2.5% significance level, can you conclude that the average credit limit for all new
credit cards issued in 2008 is higher than the corresponding average for the first 4 months of
2009?
c. Suppose that the standard deviations for the two samples were $590 and $257, respectively. Redo
parts a and b. Discuss any changes in the results.
464
Chapter 10 Estimation and Hypothesis Testing: Two Populations
10.4 Inferences About the Difference Between
Two Population Means for Paired Samples
Sections 10.1, 10.2, and 10.3 were concerned with estimation and hypothesis testing about the
difference between two population means when the two samples were drawn independently
from two different populations. This section describes estimation and hypothesis-testing pro-
cedures for the difference between two population means when the samples are dependent.
In a case of two dependent samples, two data values — one for each sample — are collected
from the same source (or element) and, hence, these are also called paired or matched samples.
For example, we may want to make inferences about the mean weight loss for members of a
health club after they have gone through an exercise program for a certain period of time. To do
so, suppose we select a sample of 15 members of this health club and record their weights be-
fore and after the program. In this example, both sets of data are collected from the same 15 per-
sons, once before and once after the program. Thus, although there are two samples, they con-
tain the same 15 persons. This is an example of paired (or dependent or matched) samples. The
procedures to make confidence intervals and test hypotheses in the case of paired samples are
different from the ones for independent samples discussed in earlier sections of this chapter.
Definition
Paired or Matched Samples Two samples are said to be paired or matched samples when for
each data value collected from one sample there is a corresponding data value collected from the
second sample, and both these data values are collected from the same source.
As another example of paired samples, suppose an agronomist wants to measure the effect
of a new brand of fertilizer on the yield of potatoes. To do so, he selects 10 pieces of land and
divides each piece into two portions. Then he randomly assigns one of the two portions from
each piece of land to grow potatoes without using fertilizer (or using some other brand of fer-
tilizer). The second portion from each piece of land is used to grow potatoes with the new brand
of fertilizer. Thus, he will have 10 pairs of data values. Then, using the procedure to be dis-
cussed in this section, he will make inferences about the difference in the mean yields of pota-
toes with and without the new fertilizer.
The question arises, why does the agronomist not choose 10 pieces of land on which to
grow potatoes without using the new brand of fertilizer and another 10 pieces of land to grow
potatoes by using the new brand of fertilizer? If he does so, the effect of the fertilizer might be
confused with the effects due to soil differences at different locations. Thus, he will not be able
to isolate the effect of the new brand of fertilizer on the yield of potatoes. Consequently, the re-
sults will not be reliable. By choosing 10 pieces of land and then dividing each of them into
two portions, the researcher decreases the possibility that the difference in the productivities of
different pieces of land affects the results.
In paired samples, the difference between the two data values for each element of the two
samples is denoted by d. This value of d is called the paired difference. We then treat all the
values of d as one sample and make inferences applying procedures similar to the ones used for
one-sample cases in Chapters 8 and 9. Note that because each source (or element) gives a pair
of values (one for each of the two data sets), each sample contains the same number of values.
That is, both samples are the same size. Therefore, we denote the (common) sample size by n,
which gives the number of paired difference values denoted by d. The degrees of freedom for
the paired samples are n — 1. Let
l± d = the mean of the paired differences for the population
<j a = the standard deviation of the paired differences for the population, which is
usually never known
d = the mean of the paired differences for the sample
s d = the standard deviation of the paired differences for the sample
n = the number of paired difference values
10.4 Inferences About the Difference Between Two Population Means for Paired Samples 465
Mean and Standard Deviation of the Paired Differences for Two Samples The values of the mean
and standard deviation, d and s d , respectively, of paired differences for two samples are calculated as 2
- Xd
d =
n
In paired samples, instead of using 3cj — x 2 as the sample statistic to make inferences about
— jj, 2 , we use the sample statistic d to make inferences about ix d . Actually the value of d is
always equal to x l — x 2 , and the value of fi d is always equal to — /jl 2 .
Sampling Distribution, Mean, and Standard Deviation of d If <r d is known and either the sample
size is large (n s 30) or the population is normally distributed, then the sampling distribution of
d is approximately normal with its mean and standard deviation given as, respectively,
MS = and o- d = —=
Vn
Thus, if the standard deviation a d of the population paired differences is known and either
the sample size is large (i.e., n & 30) or the population of paired differences is normally dis-
tributed (with n < 30), then the normal distribution can be used to make a confidence interval
and test a hypothesis about jx d . However, usually cr d is never known. Then, if the standard de-
viation a d of the population paired differences is unknown and either the sample size is large
(i.e., n s 30) or the population of paired differences is normally distributed (with n < 30), then
the t distribution is used to make a confidence interval and test a hypothesis about /j, d .
Making Inferences About fi d If
1. The standard deviation a d of the population paired differences is unknown
2. At least one of the following two conditions is fulfilled:
i. The sample size is large (i.e., n £ 30)
ii. If the sample size is small, then the population of paired differences is normally distributed
then the t distribution is used to make inferences about jx d . The standard deviation a d of d is
estimated by s d , which is calculated as
Sections 10.4.1 and 10.4.2 describe the procedures used to make a confidence interval and
test a hypothesis about jx d under the above conditions. The inferences are made using the t dis-
tribution.
10.4.1 Interval Estimation of fi d
The mean d of paired differences for paired samples is the point estimator of ix d . The follow-
ing formula is used to construct a confidence interval for jjb d when the t distribution is used.
2 The basic formula used to calculate s d is
l%(d - df
n - 1
However, we will not use this formula to make calculations in this chapter.
466 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Confidence Interval for /i rf The (1 - a) 100% confidence interval for [i d is
d ± ts d
where the value of t is obtained from the t distribution table for the given confidence level and
n — 1 degrees of freedom, and s d is calculated as explained above.
Constructing a confidence
interval for /u, rf : paired
samples, o~ d unknown, and
population normal.
Example 10-10 illustrates the procedure to construct a confidence interval for /jL d .
■ EXAMPLE 10-10
A researcher wanted to find the effect of a special diet on systolic blood pressure. She selected
a sample of seven adults and put them on this dietary plan for 3 months. The following table
gives the systolic blood pressures (in mm Hg) of these seven adults before and after the com-
pletion of this plan.
Before
210
180
195
220
231
199
224
After
193
186
186
223
220
183
233
Let i± d be the mean reduction in the systolic blood pressures due to this special dietary plan
for the population of all adults. Construct a 95% confidence interval for /j, d . Assume that the
population of paired differences is (approximately) normally distributed.
Solution Because the information obtained is from paired samples, we will make the
confidence interval for the paired difference mean (x d of the population using the paired
difference mean d of the sample. Let d be the difference in the systolic blood pressure of
an adult before and after this special dietary plan. Then, d is obtained by subtracting the
systolic blood pressure after the plan from the systolic blood pressure before the plan. The
third column of Table 10.1 lists the values of d for the seven adults. The fourth column of
the table records the values of d 2 , which are obtained by squaring each of the d values.
Table 10.1
Difference
Before
After
d
d 2
210
193
17
289
180
186
-6
36
195
186
9
81
220
223
-3
9
231
220
11
121
199
183
16
256
224
233
-9
81
Id = 35
2d 2 = 873
The values of d and s d are calculated as follows
- Xd 35
5.00
. 2 ($d) 2 I (35) 2
Xd 2 -- — - /873
10.78579312
10.4 Inferences About the Difference Between Two Population Means for Paired Samples
Hence, the standard deviation of d is
s d 10.78579312
Vn V7
4.07664661
Here, a d is not known, the sample size is small, but the population is normally distributed.
Hence, we will use the t distribution to make the confidence interval. For the 95% confidence
interval, the area in each tail of the t distribution curve is
Area in each tail = a/2 = (1 - .95)/2 = .025
The degrees of freedom are
d/=n-l=7-l=6
From the t distribution table, the t value for df = 6 and .025 area in the right tail of the t dis-
tribution curve is 2.447. Therefore, the 95% confidence interval for ii d is
d ± ts- d = 5.00 ± 2.447(4.07664661) = 5.00 ± 9.98 = -4.98 to 14.98
Thus, we can state with 95% confidence that the mean difference between systolic blood
pressures before and after the given dietary plan for all adult participants is between —4.98
and 14.98 mm Hg. ■
10.4.2 Hypothesis Testing About fi d
A hypothesis about jx d is tested by using the sample statistic d. This section illustrates the case
of the t distribution only. Earlier in this section we learned what conditions should hold true to
use the t distribution to test a hypothesis about ^ d . The following formula is used to calculate
the value of the test statistic t when testing a hypothesis about jx d .
Test Statistic t for d The value of the test statistic t for d is computed as follows:
d- H d
The critical value of t is found from the t distribution table for the given significance level and
n — 1 degrees of freedom.
Examples 10-11 and 10-12 illustrate the hypothesis-testing procedure for fjb d .
M EXAMPLE 10-11
A company wanted to know if attending a course on "how to be a successful salesperson" can
increase the average sales of its employees. The company sent six of its salespersons to attend
this course. The following table gives the 1-week sales of these salespersons before and after
they attended this course.
Before
12
18
25
9
14
16
After
18
24
24
14
19
20
Using the 1% significance level, can you conclude that the mean weekly sales for all sales-
persons increase as a result of attending this course? Assume that the population of paired dif-
ferences has a normal distribution.
Solution Because the data are for paired samples, we test a hypothesis about the paired dif-
ferences mean j± d of the population using the paired differences mean d of the sample.
Let
d = (Weekly sales before the course) — (Weekly sales after the course)
In Table 10.2, we calculate d for each of the six salespersons by subtracting the sales after the
course from the sales before the course. The fourth column of the table lists the values of d 2 .
Conducting a left-tailed
test of hypothesis about
for paired samples;
<r d not known, small
sample but normally
distributed population.
468 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Table 10.2
Before
After
Difference
d
d 2
1_
1
—
30
18
24
-6
36
25
24
1
1
9
14
-5
25
14
19
-5
25
16
20
-4
16
Xd = -25
td 2 = 139
The values of d and s d are calculated as follows:
- Zd -25
d = = = -4.17
n 6
" - V ~^r~ ■ V 6-! - 2 - miim
The standard deviation of d is
s rf 2.63944439
s 3 = -7= = 7= = 1.07754866
Vn V6
Step 1. Sfafe the null and alternative hypotheses.
We are to test if the mean weekly sales for all salespersons increase as a result of taking
the course. Let fi l be the mean weekly sales for all salespersons before the course and /jl 2 the
mean weekly sales for all salespersons after the course. Then jx d = /£, — fi 2 . The mean
weekly sales for all salespersons will increase due to attending the course if fji. l is less than
jx 2 , which can be written as — /jl 2 < or ix d < 0. Consequently, the null and alternative
hypotheses are, respectively,
H Q : /jl cI = (/J,, — j± 2 = or the mean weekly sales do not increase)
Hi. fjb d < (/a, — j± 2 < or the mean weekly sales do increase)
Note that we can also write the null hypothesis as fi d > 0.
Step 2. Select the distribution to use.
Here cr d is unknown, the sample size is small (n < 30), but the population of paired differ-
ences is normally distributed. Therefore, we use the t distribution to conduct the test.
Step 3. Determine the rejection and nonrejection regions.
The < sign in the alternative hypothesis indicates that the test is left-tailed. The signifi-
cance level is .01. Hence,
Area in left tail = a = .01
Degrees of freedom = n — 1=6 — 1=5
The critical value of t for df = 5 and .01 area in the left tail of the t distribution curve is
— 3.365. This value is shown in Figure 10.6.
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for d is computed as follows:
^ From H
10.4 Inferences About the Difference Between Two Population Means for Paired Samples 469
Step 5. Make a decision.
Because the value of the test statistic t = —3.870 for d falls in the rejection region, we re-
ject the null hypothesis. Consequently, we conclude that the mean weekly sales for all sales-
persons increase as a result of this course.
Using the p-Value to Make a Decision
We can use the p-value approach to make the above decision. To do so, we keep Steps 1 and
2 of this example. Then in Step 3, we calculate the value of the test statistic t for d (as done in
Step 4 above) and then find the /7-value for this t from the t distribution table (Table V of Ap-
pendix C) or by using technology. If we have access to technology, we can use it to find the
exact p-value, which will be .006. By using Table V, we can find the range of the p-value. From
Table V, for df = 5, the test statistic t = —3.870 falls between —3.365 and —4.032. The areas
in the t distribution table that correspond to —3.365 and —4.032 are .01 and .005, respectively.
Because it is a left-tailed test, the p-value is between .01 and .005, which can be written as
.005 < p-vafue < .01
Since we will reject the null hypothesis for any a (significance level) that is greater than the
p-value, we will reject the null hypothesis in this example for any a > .006 using the tech-
nology and a > .01 using the above range. Since a = .01 in this example, which is larger
than .006 obtained from technology, we reject the null hypothesis. Also, because a is equal
to .01, using the p-value range we reject the null hypothesis. H
EXAMPLE 10-12
Refer to Example 10-10. The table that gives the blood pressures (in mm Hg) of seven adults
before and after the completion of a special dietary plan is reproduced here.
Before
210
180
195
220
231
199
224
After
193
186
186
223
220
183
233
Let jji d be the mean of the differences between the systolic blood pressures before and after com-
pleting this special dietary plan for the population of all adults. Using the 5% significance level,
can you conclude that the mean of the paired differences jx d is different from zero? Assume that
the population of paired differences is (approximately) normally distributed.
Solution Table 10.3 gives d and d 2 for each of the seven adults (blood pressure values in
mm Hg).
The values of d and s d are calculated as follows:
1 _ ^ d _ 35
n ~ 7
Xd 2
= 5.00
Wf
873
(35) 2
n - 1
10.78579312
Making a two-tailed
test of hypothesis about /x rf for
paired samples: cr d not known,
small sample but normally
distributed population.
470 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Table 10.3
Difference
Rpffti*p
After
el
d 2
210
193
17
289
180
186
-6
36
195
186
9
81
220
223
-3
9
231
220
11
121
199
183
16
256
224
233
-9
81
Xd = 35
td 2 = 873
Hence, the standard deviation of d is
s d 10.78579312
*3
4.07664661
Vn V7
Step 1. State the null and alternative hypotheses.
H : jju d = (The mean of the paired differences is not different from zero.)
H x : i± d ¥= (The mean of the paired differences is different from zero.)
Step 2. Select the distribution to use.
Here a d is unknown, the sample size is small, but the population of paired differences is
(approximately) normal. Hence, we use the t distribution to make the test.
Step 3. Determine the rejection and nonrejection regions.
The # sign in the alternative hypothesis indicates that the test is two-tailed. The signifi-
cance level is .05.
Area in each tail of the curve = a/2 = .05/2 = .025
Degrees of freedom = n — 1=7 — 1=6
The two critical values of t for df = 6 and .025 area in each tail of the t distribution curve are
—2.447 and 2.447. These values are shown in Figure 10.7.
Figure 10.7
Reject H
a/2 = .025
■ Do not reject H
Reject H
a/2 = .025
-2.447 2.447
-Two critical values of t-
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for d is computed as follows:
d - n d 5.00 -
From Hn
4.07664661
1.226
Step 5. Make a decision.
Because the value of the test statistic t = 1.226 for d falls in the nonrejection region, we
fail to reject the null hypothesis. Hence, we conclude that the mean of the population paired
10.4 Inferences About the Difference Between Two Population Means for Paired Samples 471
differences is not different from zero. In other words, we can state that the mean of the dif-
ferences between the systolic blood pressures before and after completing this special dietary
plan for the population of all adults is not different from zero.
Using the p-Value to Make a Decision
We can use the /?-value approach to make the above decision. To do so, we keep Steps 1 and
2 of this example. Then in Step 3, we calculate the value of the test statistic t for d (as done
in Step 4 above) and then find the /7-value for this t from the t distribution table (Table V of
Appendix C) or by using technology. If we have access to technology, we can use it to find
the exact /?-value, which will be .266. By using Table V, we can find the range of the /7-value.
From Table V, for df = 6, the test statistic t = 1 .226 is less than 1 .440. The area in the t dis-
tribution table that corresponds to 1.440 is .10. Because it is a two-tailed test, the /?-value is
greater than 2(.10) = .20, which can be written as
/5-value > .20
Since a = .05 in this example, which is smaller than .20 and also .266 (obtained from tech-
nology), we fail to reject the null hypothesis. I
EXERCISES
CONCEPTS AND PROCEDURES
10.47 Explain when you would use the paired-samples procedure to make confidence intervals and test
hypotheses.
10.48 Find the following confidence intervals for n lh assuming that the populations of paired differences
are normally distributed.
a. n = 11, d = 25.4, s d = 13.5, confidence level = 99%
b. n = 23, d = 13.2,
= 4.8, confidence level = 95%
c. n =18, d = 34.6, s d = 11.7, confidence level = 90%
10.49 Find the following confidence intervals for n lh assuming that the populations of paired differences
are normally distributed.
a. n =12, d = 17.5, s d = 6.3, confidence level = 99%
b. n = 27, d\ = 55.9, s d = 14.7, confidence level = 95%
c. n = 16, d = 29.3, s d = 8.3, confidence level = 90%
10.50 Perform the following tests of hypotheses, assuming that the populations of paired differences are
normally distributed.
a. H : /x rf = 0, H x : /jl cI + 0, n = 9, d = 6.7, s d = 2.5, a =.10
b. H : = 0, fj. d > 0, n = 22, d = 14.8, s d = 6.4, a = .05
c. H : iL d = 0, H { . fi d < 0,
17,
-9.3,
4.8,
.01
10.51 Conduct the following tests of hypotheses, assuming that the populations of paired differences are
normally distributed.
a. H : fi d = 0, H x : jx d ^0, n = 26, d = 9.6, s d = 3.9, a = .05
b. H : fi d = 0, H t : ft, > 0,
c. H : iL d = 0, ffj: fi d < 0,
15,
20,
-7.4,
4.7,
2.3,
a
a
.01
.10
■ APPLICATIONS
10.52 A company sent seven of its employees to attend a course in building self-confidence. These em-
ployees were evaluated for their self-confidence before and after attending this course. The following table
gives the scores (on a scale of 1 to 15, 1 being the lowest and 15 being the highest score) of these em-
ployees before and after they attended the course.
Before
8
5
4
9
6
9
5
After
10
8
5
11
6
7
9
472 Chapter 10 Estimation and Hypothesis Testing: Two Populations
a. Construct a 95% confidence interval for the mean jx d of the population paired differences, where
a paired difference is equal to the score of an employee before attending the course minus the
score of the same employee after attending the course.
b. Test at the 1% significance level whether attending this course increases the mean score of em-
ployees.
Assume that the population of paired differences has a normal distribution.
10.53 Several retired bicycle racers are coaching a large group of young prospects. They randomly select seven
of their riders to take part in a test of the effectiveness of a new dietary supplement that is supposed to increase
strength and stamina. Each of the seven riders does a time trial on the same course. Then they all take the di-
etary supplement for 4 weeks. All other aspects of their training program remain as they were prior to the time
trial. At the end of the 4 weeks, these riders do another time trial on the same course. The times (in minutes)
recorded by each rider for these trials before and after the 4-week period are shown in the following table.
Before
103
97
111
95
102
96
108
After
100
95
104
101
96
91
101
Construct a 99% confidence interval for the mean fi d of the population paired differences, where
a paired difference is equal to the time taken before the dietary supplement minus the time taken
after the dietary supplement.
Test at the 2.5% significance level whether taking this dietary supplement results in faster times
in the time trials.
Assume that the population of paired differences is (approximately) normally distributed.
10.54 One type of experiment that might be performed by an exercise physiologist is as follows: Each
person in a random sample is tested in a weight room to determine the heaviest weight with which he or
she can perform an incline press five times with his or her dominant arm (defined as the hand that a per-
son uses for writing). After a significant rest period, the same weight is determined for each individual's
nondominant arm. The physiologist is interested in the differences in the weights pressed by each arm.
The following data represent the maximum weights (in pounds) pressed by each arm for a random sam-
ple of 18 fifteen-year old girls. Assume that the differences in weights pressed by each arm for all fifteen-
year old girls are approximately normally distributed.
Subject
Dominant
Arm
Nondominant
Arm
Subject
Dominant
Arm
Nondominant
Arm
1
59
53
10
47
38
2
32
30
11
40
35
3
27
24
12
36
36
4
18
20
13
21
25
5
42
40
14
51
48
6
12
12
15
30
30
7
29
24
16
32
31
8
33
34
17
14
14
9
22
22
18
26
27
a. Make a 99% confidence interval for the mean of the paired differences for the two populations,
where a paired difference is equal to the maximum weight for the dominant arm minus the max-
imum weight for the nondominant arm.
b. Using the 2% significance level, can you conclude that the average paired difference as defined
in part a is positive?
10.55 As was mentioned in Exercise 9.38, The Bath Heritage Days, which take place in Bath, Maine, switched
to a Whoopie Pie eating contest in 2009. Suppose the contest involves eating nine Whoopie Pies, each weigh-
ing 1/3 pound. The following data represent the times (in seconds) taken by each of the 13 contestants (all
of whom finished all nine Whoopie Pies) to eat the first Whoopie Pie and the last (ninth) Whoopie Pie.
Contestant
1
2
3
4
5
6
7
8
9
10
11
12
13
First pie
49
59
66
49
63
70
77
59
64
69
60
58
71
Last pie
49
74
92
93
91
73
103
59
85
94
84
87
111
10.5 Inferences About the Difference Between Two Population Proportions for Large and Independent Samples 473
a. Make a 95% confidence interval for the mean of the population paired differences, where a paired
difference is equal to the time taken to eat the ninth pie (which is the last pie) minus the time
taken to eat the first pie.
b. Using the 10% significance level, can you conclude that the average time taken to eat the ninth pie
(which is the last pie) is at least 15 seconds more than the average time taken to eat the first pie.
Assume that the population of paired differences is (approximately) normally distributed.
10.56 The manufacturer of a gasoline additive claims that the use of this additive increases gasoline
mileage. A random sample of six cars was selected, and these cars were driven for 1 week without the
gasoline additive and then for 1 week with the gasoline additive. The following table gives the miles per
gallon for these cars without and with the gasoline additive.
Without
24.6
28.3
18.9
23.7
15.4 29.5
With
26.3
31.7
18.2
25.3
18.3 30.9
a. Construct a 99% confidence interval for the mean of the population paired differences, where
a paired difference is equal to the miles per gallon without the gasoline additive minus the miles
per gallon with the gasoline additive.
b. Using the 2.5% significance level, can you conclude that the use of the gasoline additive in-
creases the gasoline mileage?
Assume that the population of paired differences is (approximately) normally distributed.
10.57 A factory that emits airborne pollutants is testing two different brands of filters for its smokestacks.
The factory has two smokestacks. One brand of filter (Filter I) is placed on one smokestack, and the other
brand (Filter II) is placed on the second smokestack. Random samples of air released from the smoke-
stacks are taken at different times throughout the day. Pollutant concentrations are measured from both
stacks at the same time. The following data represent the pollutant concentrations (in parts per million)
for samples taken at 20 different times after passing through the filters. Assume that the differences in
concentration levels at all times are approximately normally distributed.
Time
Filter I
Filter II
Time
Filter I
Filter II
1
24
26
11
11
9
2
31
30
12
8
10
3
35
33
13
14
17
4
32
28
14
17
16
5
25
23
15
19
16
6
25
28
16
19
18
7
29
24
17
25
27
8
30
33
18
20
22
9
26
22
19
23
27
10
18
18
20
32
31
a. Make a 95% confidence interval for the mean of the population paired differences, where a paired
difference is equal to the pollutant concentration passing through Filter I minus the pollutant
concentration passing through Filter II.
b. Using the 5% significance level, can you conclude that the average paired difference for con-
centration levels is different from zero?
10.5 Inferences About the Difference Between
Two Population Proportions for Large and
Independent Samples
Quite often we need to construct a confidence interval and test a hypothesis about the differ-
ence between two population proportions. For instance, we may want to estimate the difference
between the proportions of defective items produced on two different machines. If p l and p 2 are
the proportions of defective items produced on the first and second machine, respectively, then
474 Chapter 10 Estimation and Hypothesis Testing: Two Populations
we are to make a confidence interval for p x — p 2 . Alternatively, we may want to test the hy-
pothesis that the proportion of defective items produced on Machine I is different from the pro-
portion of defective items produced on Machine II. In this case, we are to test the null hypoth-
esis p x — p 2 = against the alternative hypothesis p l — p 2 ¥= 0.
This section discusses how to make a confidence interval and test a hypothesis about p l — p 2
for two large and independent samples. The sample statistic that is used to make inferences
about p x — p 2 is /?, — p 2 , where p x and p 2 are the proportions for two large and independent
samples. As discussed in Chapter 7, we determine a sample proportion by dividing the number
of elements in the sample that possess a given attribute by the sample size. Thus,
where x x and x 2 are the number of elements that possess a given characteristic in the two sam-
ples and n, and n 2 are the sizes of the two samples, respectively.
10.5.1 Mean, Standard Deviation, and Sampling
Distribution of p, - p 2
As discussed in Chapter 7, for a large sample, the sample proportion p is (approximately) nor-
mally distributed with mean p and standard deviation \fpqjn. Hence, for two large and inde-
pendent samples of sizes n x and n 2 , respectively, their sample proportions p x and p 2 are
(approximately) normally distributed with means p x and p 2 and standard deviations \fp x q x /n x
and \ip 2 q 2 /n 2 , respectively. Using these results, we can make the following statements about
the shape of the sampling distribution of p x — p 2 and its mean and standard deviation.
Mean, Standard Deviation, and Sampling Distribution of p, - p 2 For two large and independent
samples, the sampling distribution of p x — p 2 is (approximately) normal, with its mean and stan-
dard deviation given as
respectively, where q x = 1 — p x and q 2 = 1 — p 2 .
Thus, to construct a confidence interval and test a hypothesis about p x — p 2 for large and
independent samples, we use the normal distribution. As was indicated in Chapter 7, in the case
of proportion, the sample is large if np and nq are both greater than 5. In the case of two sam-
ples, both sample sizes are large if n x p x , n x q x , n 2 p 2 , and n 2 q 2 are all greater than 5.
10.5.2 Interval Estimation of p, - p 2
The difference between two sample proportions p x — p 2 is the point estimator for the difference
between two population proportions p x — p 2 . Because we do not know p x and p 2 when we are
making a confidence interval for p x — p 2 , we cannot calculate the value of <r& -p . Therefore,
we use i/;,-,?, as the point estimator of in the interval estimation. We construct the con-
fidence interval for p x — p 2 using the following formula.
Confidence Interval for p, - p 2 The (1 - a) 100% confidence interval for p x - p 2 is
(Pl ~ Pl) ± ZSft-p 2
where the value of z is read from the normal distribution table for the given confidence level,
and spt-pt is calculated as
p x = x x /n x and p 2 = x 2 /n 2
Wi-h = Pi ~ Pi
and
10.5 Inferences About the Difference Between Two Population Proportions for Large and Independent Samples 475
Example 10-13 describes the procedure used to make a confidence interval for the differ-
ence between two population proportions for large samples.
■ EXAMPLE 10-13
A researcher wanted to estimate the difference between the percentages of users of two tooth-
pastes who will never switch to another toothpaste. In a sample of 500 users of Toothpaste A
taken by this researcher, 100 said that they will never switch to another toothpaste. In another
sample of 400 users of Toothpaste B taken by the same researcher, 68 said that they will never
switch to another toothpaste.
(a) Let pj and p 2 be the proportions of all users of Toothpastes A and B, respectively, who
will never switch to another toothpaste. What is the point estimate of p x — p 2 l
(b) Construct a 97% confidence interval for the difference between the proportions of all
users of the two toothpastes who will never switch.
Solution Let p x and p 2 be the proportions of all users of Toothpastes A and B, respectively,
who will never switch to another toothpaste, and let p x and p 2 be the respective sample pro-
portions. Let x x and x 2 be the number of users of Toothpastes A and B, respectively, in the
two samples who said that they will never switch to another toothpaste. From the given
information,
Toothpaste A : n x = 500 and x x = 100
ToothpasteB : n 2 = 400 and x 2 = 68
The two sample proportions are calculated as follows:
Pi = xjn x = 100/500
p 2 = x 2 /n 2 = 68/400
Then,
$! = 1 - .20 = .80 and q 2 = 1 - .17 = .83
(a) The point estimate of p x — p 2 is as follows:
Point estimate of p x — p 2 = p x — p 2 = .20 — .17 = .03
(b) The values of nip x , n x q x , n 2 p 2 , and n 2 q 2 are
njpi = 500(.20) = 100 n x q x = 500(.80) = 400
n 2 p 2 = 400(.17) = 68 n 2 q 2 = 400(.83) = 332
Because each of these values is greater than 5, both sample sizes are large.
Consequently we use the normal distribution to make a confidence interval for/?, — p 2 .
The standard deviation of p x — p 2 is
_ jEk^hk . mw^imm _ 02593742
'" Pl V Hi n 2 V 500 400
The z value for a 97% confidence level, obtained from the normal distribution table
is 2.17. The 97% confidence interval for p x — p 2 is
(p 1 -p 2 ) ± zs Pl -p, = (.20 - .17) ± 2.17(.02593742)
= .03 ± .056 = -.026 to .086
Thus, with 97% confidence we can state that the difference between the two popula-
tion proportions is between —.026 and .086.
Note that here p x — p 2 = .03 gives the point estimate of p x — p 2 and
zsp { _^ = .056 is the margin of error of the estimate. H
476 Chapter 10 Estimation and Hypothesis Testing: Two Populations
10.5.3 Hypothesis Testing About p, - p 2
In this section we learn how to test a hypothesis about p x — p 2 for two large and independent
samples. The procedure involves the same five steps we have used previously. Once again, we
calculate the standard deviation of /3, — p 2 as
/Plgl Pill
V «i n 2
When a test of hypothesis about p x — p 2 is performed, usually the null hypothesis is p x = p 2 and
the values of p x and p 2 are not known. Assuming that the null hypothesis is true and p x = p 2 , a com-
mon value of p x and p 2 , denoted by p, is calculated by using one of the following two formulas:
_ x x + x-i n x p x + n 2 p 2
p = — — — or
77, + n 2 n x + n 2
Which of these formulas is used depends on whether the values of x x and x 2 or the values of p x
and p 2 are known. Note that x x and x 2 are the number of elements in each of the two samples
that possess a certain characteristic. This value of p is called the pooled sample proportion.
Using the value of the pooled sample proportion, we compute an estimate of the standard de-
viation of p x — p 2 as follows:
'A-A = VKnt + £)
where q = 1 — p.
Test Statistic 1 for Pi — p 2 The value of the test statistic zfor p x — p 2 is calculated as
(Pi ~ Pi) ~ (Pi ~ Pi)
z =
S P<-P2
The value of p x — p 2 is substituted from H , which usually is zero.
Examples 10-14 and 10-15 illustrate the procedure to test hypotheses about the difference
between two population proportions for large samples.
■ EXAMPLE 10-14
Reconsider Example 10-13 about the percentages of users of two toothpastes who will never
switch to another toothpaste. At the 1% significance level, can you conclude that the propor-
tion of users of Toothpaste A who will never switch to another toothpaste is higher than the
proportion of users of Toothpaste B who will never switch to another toothpaste?
Solution Let p x and p 2 be the proportions of all users of Toothpastes A and B, respectively,
who will never switch to another toothpaste, and let p x and p 2 be the corresponding sample pro-
portions. Let x x and x 2 be the number of users of Toothpastes A and B, respectively, in the two
samples who said that they will never switch to another toothpaste. From the given information,
ToothpasteA : n x = 500 and x x = 100
ToothpasteB : n 2 = 400 and x 2 = 68
The significance level is a = .01. The two sample proportions are calculated as follows:
p x = x x /n x = 100/500 = .20
p 2 = x 2 /n 2 = 68/400 = .17
Step 1. State the null and alternative hypotheses.
We are to test if the proportion of users of Toothpaste A who will never switch to another
toothpaste is higher than the proportion of users of Toothpaste B who will never switch to
Making a right-tailed test of
hypothesis about p i — p,: large
and independent samples.
10.5 Inferences About the Difference Between Two Population Proportions for Large and Independent Samples 477
another toothpaste. In other words, we are to test whether p x is greater than p 2 . This can be
written as p x — p 2 > 0. Thus, the two hypotheses are
H : p x = p 2 or p x — p 2 = (p x is not greater than p 2 )
Hv Pi ^ Pi or P\ ~ Pi > (Pi i s greater than p 2 )
Step 2. Select the distribution to use.
As shown in Example 10-13, n^pi, n 2 p 2 , and n 2 q 2 are all greater than 5. Consequently
both samples are large, and we use the normal distribution to make the test.
Step 3. Determine the rejection and nonrejection regions.
The > sign in the alternative hypothesis indicates that the test is right-tailed. From the nor-
mal distribution table, for a .01 significance level, the critical value of z is 2.33 for .9900 area
to the left. This is shown in Figure 10.8.
Figure 10.8
r a= .01
Pi -Pz =
A A
Pl ~ P2
Do not reject H — »-
Reject H Q
Critical value of z
2.33
Step 4. Calculate the value of the test statistic
The pooled sample proportion is
_ x x + x 2 100 + 68
.187
and
n x + n 2 500 + 400
The estimate of the standard deviation of p x — p 2 is
1 - .187
\ — + —
"l "2
(.187)(.813)( — + — )
A 500 400/
.813
The value of the test statistic z for p x — p 2 is
(Pi-P2)-(Pi-P2) = (-20--17)-0
s s _ s „ .02615606
.02615606
■ From H
1.15
Step 5. Make a decision.
Because the value of the test statistic z = 1.15 for p x — p 2 falls in the nonrejection region,
we fail to reject the null hypothesis. Therefore, we conclude that the proportion of users of
Toothpaste A who will never switch to another toothpaste is not greater than the proportion
of users of Toothpaste B who will never switch to another toothpaste.
Using the p-Value to Make a Decision
We can use the ;?-value approach to make the above decision. To do so, we keep Steps 1 and
2 above. Then in Step 3, we calculate the value of the test statistic z (as done in Step 4 above)
and find the ;?-value for this z from the normal distribution table. In Step 4 above, the z-value
for p x — p 2 was calculated to be 1.15. In this example, the test is right-tailed. The /7-value is
given by the area under the normal distribution curve to the right of z = 1.15. From the nor-
mal distribution table (Table IV of Appendix C), this area is 1 — .8749 = .1251. Hence, the
IS VACATION
IMPORTANT?
USA TODAY Snapshots®
Vacation more
important for
younger generation
Percentage who said
caking a vacation is very
important, by age:
6 Y*
Source: Access America and Ipsos Public Affairs survey or 1,000
dtlulls 1 S din] ultier. Mar Kin i>F t?i iuj 23 ui?i c en MKe utiliu*
By J ae Yang and S am Ward, USA TODay
The above chart shows the percentage of adults in three different age groups who say that taking a vaca-
tion is very important for them. These results are based on a survey of 1000 adults 18 and older. Accord-
ing to this survey, 74% of adults in the age group 18 to 34 said that taking a vacation is very important to
them. The corresponding percentages for the 35 to 54 years age group and 55 years and older group were
66% and 61%, respectively. If we know the sample sizes for these three groups included in this survey, we
can make a confidence interval and perform a test of hypothesis for the difference in the percentages for
any two groups listed in the chart using the procedures learned in this section.
Suppose that of the 1000 adults included in the survey, 400 were 18 to 34 years old, 320 were 35 to
54 years old, and 280 were 55 years and older. Consider the first two groups, 18 to 34 years old and 35
to 54 years old. Let us call them Group 1 and Group 2, respectively. Let p, and p 2 be the proportion of
adults in group 1 and Group 2, respectively, who would say that vacations are very important for them. Let
p, and p 2 be the corresponding sample proportions. Then, from the given information:
For Group 1 : n, = 400 p, = .74 c), = 1 - .74 = .26
For Group 2: n 2 = 320 p 2 = .66 q 2 = 1 - .66 = .34
Below we make a confidence interval and test a hypothesis about p, - p 2 for these two groups.
1 . Confidence interval for p, - p 2
Suppose we want to make a 97% confidence interval for p, - p 2 . The z value from Table IV for the 97%
confidence level is 2.17. The standard deviation of p, - p 2 is
[M\ ~ PA / (.74)(.26) (.66X34) _ _ _
Ss -s = \ 1 = \/ 1 = .03438386
Pl P! V n, n 2 V 400 320
p-value is .1251. We reject the null hypothesis for any a (significance level) greater than the
p-value; in this example, we will reject the null hypothesis for any a > .1251 or 12.51%. Be-
cause a = .01 here, which is less than .1251, we fail to reject the null hypothesis. H
Conducting a two-tailed test of
hypothesis about p t — p 2 : large
and independent samples.
■ EXAMPLE 10-15
According to a July 1, 2009, Quinnipiac University poll, 62% of adults aged 18 to 34 years
and 50% of adults aged 35 years and older surveyed believed that it is the government's re-
sponsibility to make sure that everyone in the United States has adequate health care
(www.quinnipiac.edu/xl295. xml?ReleaseID=1344). The survey included approximately 683
people in the 18- to 34-year age group and 2380 people aged 35 years and older. Test whether
the proportions of people who believe that it is the government's responsibility to make sure
478
Hence, the 97% confidence interval for p, - p 2 is:
(pi ~ Pi) ± z Sp
(.74 - .66) ± 2.17(.03438386)
.005 to .1 55 or .5% to 1 5.5%
.08
.075
Thus, we can say with 97% confidence that the difference in the proportions of all adults in Group 1 and
Group 2 who feel vacation is important for them is in the interval .005 to .155 or .5% to 15.5%.
2. Test of hypothesis about p, - p 2
Suppose we want to test, at the 1% significance level, whether the proportion of all adults in the age group
18 to 34 years who say vacation is very important to them is greater than that for all adults in the age group
35 to 54 years. In other words, we are to test if p, is greater than p 2 . The null and alternative hypotheses are
H : p, = p 2 or p, - p 2 =
H,:p, >p 2 or p, -p 2 >
Note that the test is right-tailed. For a = .01, the critical value of z from the normal distribution table for
.9900 is 2.33. Thus, we will reject the null hypothesis if the observed value of z is 2.33 or larger. The pooled
sample proportion is
n,p, + n 2 p 2 400(.74) + 320(.66)
P n, + n 2 400 + 320
and q = 1 - p = 1 - .704 = .296
The estimate of the standard deviation of p, - p 2 is
^£ + i)S ( - 704)( - 2 ^400 + 32o)
The value of the test statistic z for p, - p 2 is
.704
.03423682
(Pi - Pi) ~ (Pi - Pi)
(.74 - .66) -
.03423682
2.34
Since the observed value of z = 2.34 is larger than the critical value of z = 2.33, we reject the null hypoth-
esis. As a result we conclude that p, is greater than p 2 , and that the proportion of all adults in the age
group 18 to 34 years who will say that vacation is important is greater than that for all adults in the age
group 35 to 54 years. Note that the observed value of z is very close to the critical value of z and, hence,
the conclusion is not strong. By decreasing the significance level slightly, we can change the conclusion.
We can also use the p-value approach to make this decision. In this example, the test is right-tailed.
As calculated above, the z value for p, - p 2 is 2.34. From the normal distribution table, the area to the right
of z = 2.34 is 1 - .9904 = .0096. Hence, the p-value is .0096. Since, a = .01 in this example is greater
than .0096, we reject the null hypothesis and conclude that the proportion of all adults in the age group
18 to 34 years who will say that vacation is very important is greater than that for all adults in the age
group 35 to 54 years.
Source: The chart reproduced with
permission from USA TODAY, August
31, 2009. Copyright © 2009, USA
TODAY.
that everyone in the United States has adequate health care are different for the two age groups.
Use a 1% significance level.
Solution Let p l and p 2 be the proportions of all adults in the two age groups — 18- to 34-year
age group and 35 years and older group, respectively — who believe that it is the government's
responsibility to make sure that everyone in the United States has adequate health care. Let p x
and p 2 be the corresponding sample proportions. From the given information,
For 1 8-34 age group: n l = 683 and p r = .62
For 35 and older group: n 2 = 2380 and p 2 = .50
The significance level is a = .01.
479
480 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Step 1. State the null and alternative hypotheses.
The null and alternative hypotheses are, respectively,
H : p x — p 2 = (The two population proportions are not different.)
Hi. pi — p 2 # (The two population proportions are different.)
Step 2. Select the distribution to use.
Because the samples are large and independent, we apply the normal distribution to make
the test. (The reader should check that «]pi, n x q x , n 2 p 2 , and n 2 q 2 are all greater than 5.)
Step 3. Determine the rejection and nonrejection regions.
The # sign in the alternative hypothesis indicates that the test is two-tailed. For a 1%
significance level, the critical values of z are —2.58 and 2.58. Note that to find these two crit-
ical values, we look for .0050 and .9950 areas in Table IV of Appendix C. These values are
shown in Figure 10.9.
Step 4. Calculate the value of the test statistic.
The pooled sample proportion is
+ n 2 p 2 683(.62) + 2380(.50)
p = = = .527
F n, + n 2 683 + 2380
and
q = 1 - p = 1 - .527 = .473
The estimate of the standard deviation of p x — p 2 is
K£ + £) = V ( - 527)( - 473) (^3- + 2^o) = -° 2167258
The value of the test statistic z for p x — p 2 is
(p! ~ Pi) ~ (Pi - Pz) (.62 -.50)-0
z = = = 5.54
s h _ h .02167258
Step 5. Make a decision.
Because the value of the test statistic z = 5.54 falls in the rejection region, we reject the null hy-
pothesis H . Therefore, we conclude that the proportions of adults in the two age groups — age group
18 to 34 years and age group 35 years and older — who believe that it is the government's respon-
sibility to make sure that everyone in the United States has adequate health care are different.
Using the p-Value to Make a Decision
We can use the p-value approach to make the above decision. To do so, we keep Steps 1 and
2 above. Then in Step 3, we calculate the value of the test statistic z (as done in Step 4 above)
10.5 Inferences About the Difference Between Two Population Proportions for Large and Independent Samples
481
and find the p-value for this z from the normal distribution table. In Step 4 above, the z-value
for p x — p 2 was calculated to be 5.54. In this example, the test is two-tailed. The /7-value is
given by twice the area under the normal distribution curve to the right of z = 5.54. From the
normal distribution table (Table IV), the area to the right of z = 5.54 is (approximately) zero.
Hence, the p--value is 2(.0000) =.0000. As we know, we will reject the null hypothesis for any
a (significance level) greater than the p-value. Since a = .01 in this example, which is greater
than .0000, we reject the null hypothesis. I
EXERCISES
CONCEPTS AND PROCEDURES
10.58 What is the shape of the sampling distribution of p, — p 2 for two large samples? What are the mean
and standard deviation of this sampling distribution?
10.59 When are the samples considered large enough for the sampling distribution of the difference
between two sample proportions to be (approximately) normal?
10.60 Construct a 99% confidence interval for p, — p 2 for the following.
n, = 300 p, = .55 n 2 = 200 p 2 = .62
10.61 Construct a 95% confidence interval forpj — p 2 for the following.
«, = 100 p, = .81 n 2 = 150 p 2 = .77
10.62 Refer to the information given in Exercise 10.60. Test at the 1% significance level if the two
population proportions are different.
10.63 Refer to the information given in Exercise 10.61. Test at the 5% significance level if p x — p 2 is
different from zero.
10.64 Refer to the information given in Exercise 10.60. Test at the 1% significance level if p x is less than p 2 .
10.65 Refer to the information given in Exercise 10.61. Test at the 2% significance level if p x is greater
than p 2 .
10.66 A sample of 500 observations taken from the first population gave X\ = 305. Another sample of
600 observations taken from the second population gave x 2 = 348.
a. Find the point estimate of p x — p 2 .
b. Make a 97% confidence interval for p x — p 2 .
c. Show the rejection and nonrejection regions on the sampling distribution of p x — p 2 for
Ho- Pi = Pi versus//!: p y > p 2 . Use a significance level of 2.5%.
d. Find the value of the test statistic z for the test of part c.
e. Will you reject the null hypothesis mentioned in part c at a significance level of 2.5%?
10.67 A sample of 1000 observations taken from the first population gave x, = 290. Another sample of
1200 observations taken from the second population gave x 2 = 396.
a. Find the point estimate of p, — p 2 .
b. Make a 98% confidence interval forp, — p 2 .
c. Show the rejection and nonrejection regions on the sampling distribution of p x — p 2 for
Hq- Pi = p 2 versus H{. p x < p 2 . Use a significance level of 1%.
d. Find the value of the test statistic z for the test of part c.
e. Will you reject the null hypothesis mentioned in part c at a significance level of 1%?
■ APPLICATIONS
10.68 According to a June 2009 report (http://www.alertnet.org/thenews/newsdesk/L31011082.htm), 68% of
people with "green" jobs in North America felt that they had job security, whereas 60% of people with green
jobs in the United Kingdom felt that they had job security. Suppose that these results were based on samples
of 305 people with green jobs from North America and 280 people with green jobs from the United Kingdom.
a. Construct a 96% confidence interval for the difference between the two population proportions.
b. Using the 2% significance level, can you conclude that the proportion of all people with green
jobs in North America who feel that they have job security is higher than the corresponding pro-
portion for the United Kingdom? Use the critical-value approach.
c. Repeat part b using the p-value approach.
482
Chapter 10 Estimation and Hypothesis Testing: Two Populations
10.69 A study in the July 7, 2009, issue of USA TODAY stated that the 401(k) participation rate among
U.S. employees of Asian heritage is 76%, whereas the participation rate among U.S. employees of His-
panic heritage is 66%. Suppose that these results were based on random samples of 100 U.S. employees
from each group.
a. Construct a 95% confidence interval for the difference between the two population proportions.
b. Using the 5% significance level, can you conclude that the 401(k) participation rates are differ-
ent for all U.S. employees of Asian heritage and all U.S. employees of Hispanic heritage? Use
the critical-value and p- value approaches.
c. Repeat parts a and b for both sample sizes of 200 instead of 100. Does your conclusion change
in part b?
10.70 A July 2009 Pew Research Center survey asked a variety of science questions of independent
random samples of scientists and the public at-large (http://people-press.org/report/528/). One of the ques-
tions asked was whether all parents should be required to vaccinate their children. The percentage of peo-
ple answering "yes" to this question was 69% of the general public and 82% of scientists. Suppose that
the survey included 110 members of the general public and 105 scientists.
a. Construct a 98% confidence interval for the difference between the two population proportions.
b. Using the 1% significance level, can you conclude that the percentage of the general public who
feels that all parents should be required to vaccinate their children is less than the percentage of
all scientists who feels that all parents should be required to vaccinate their children? Use the
critical- value and p- value approaches.
c. The actual sample sizes used in the survey were 2001 members of the general public and 1005
scientists. Repeat parts a and b using the actual sample sizes. Does your conclusion change in
part b?
10.71 A state that requires periodic emission tests of cars operates two emissions test stations, A and B,
in one of its towns. Car owners have complained of lack of uniformity of procedures at the two stations,
resulting in different failure rates. A sample of 400 cars at Station A showed that 53 of those failed the
test; a sample of 470 cars at Station B found that 51 of those failed the test.
a. What is the point estimate of the difference between the two population proportions?
b. Construct a 95% confidence interval for the difference between the two population proportions.
c. Testing at the 5% significance level, can you conclude that the two population proportions are
different? Use both the critical-value and the p-value approaches.
10.72 The management of a supermarket chain wanted to investigate if the percentages of men and
women who prefer to buy national brand products over the store brand products are different. A sam-
ple of 600 men shoppers at the company's supermarkets showed that 246 of them prefer to buy national
brand products over the store brand products. Another sample of 700 women shoppers at the company's
supermarkets showed that 266 of them prefer to buy national brand products over the store brand
products.
a. What is the point estimate of the difference between the two population proportions?
b. Construct a 95% confidence interval for the difference between the proportions of all men and
all women shoppers at these supermarkets who prefer to buy national brand products over the
store brand products.
c. Testing at the 5% significance level, can you conclude that the proportions of all men and all
women shoppers at these supermarkets who prefer to buy national brand products over the store
brand products are different?
10.73 The lottery commissioner's office in a state wanted to find if the percentages of men and women who
play the lottery often are different. A sample of 500 men taken by the commissioner's office showed that
160 of them play the lottery often. Another sample of 300 women showed that 66 of them play the lottery
often.
a. What is the point estimate of the difference between the two population proportions?
b. Construct a 99% confidence interval for the difference between the proportions of all men and
all women who play the lottery often.
c. Testing at the 1% significance level, can you conclude that the proportions of all men and all
women who play the lottery often are different?
10.74 A mail-order company has two warehouses, one on the West Coast and the second on the East
Coast. The company's policy is to mail all orders placed with it within 72 hours. The company's quality
control department checks quite often whether or not this policy is maintained at the two warehouses. A
recently taken sample of 400 orders placed with the warehouse on the West Coast showed that 364 of
them were mailed within 72 hours. Another sample of 300 orders placed with the warehouse on the East
Coast showed that 279 of them were mailed within 72 hours.
Supplementary Exercises 483
a. Construct a 97% confidence interval for the difference between the proportions of all orders
placed at the two warehouses that are mailed within 72 hours.
b. Using the 2.5% significance level, can you conclude that the proportion of all orders placed at
the warehouse on the West Coast that are mailed within 72 hours is lower than the correspon-
ding proportion for the warehouse on the East Coast?
10.75 A company that has many department stores in the southern states wanted to find at two such stores
the percentage of sales for which at least one of the items was returned. A sample of 800 sales randomly
selected from Store A showed that for 280 of them at least one item was returned. Another sample of 900
sales randomly selected from Store B showed that for 279 of them at least one item was returned.
a. Construct a 98% confidence interval for the difference between the proportions of all sales at
the two stores for which at least one item is returned.
b. Using the 1% significance level, can you conclude that the proportions of all sales for which at
least one item is returned is higher for Store A than for Store B?
USES AND MISUSES... CRAPES TO GE
Turn on the TV one Sunday morning to one of the news programs
in which journalists and writers spar with politicians. It is very com-
mon on these programs for someone to claim that events of the pre-
vious week are reminiscent of events of the previous decade and that
particular actions are thus warranted. It is just as common for a par-
ticipant in the debate to state that the interpretation is incorrect be-
cause the comparison is not "apples to apples, oranges to oranges."
Statistical analysis of the differences between two populations re-
quires that the populations be essentially similar because the meth-
ods described in the chapter are for making apples-to-apples and
oranges-to-oranges comparisons. The only possible difference be-
tween the two populations should be the characteristic that you are
studying; remember that the null hypothesis is often that there is no
difference between the population means or proportions.
To illustrate, consider an extreme example— wine. The text de-
scribes an experiment in which a statistician instructs a farmer to
plant two varieties of a crop on 20 distributed plots to ensure that
comparisons of the productivity are not affected by peculiarities of
the land on which the crops are planted. Winemakers rarely do this
because the peculiarities of the land are what they want to empha-
size. Suppose that Winemaker A and Winemaker B live in the same
region and purchase their vines from the same nursery. Because they
plant the same varietals, live near one another, experience the same
weather, use similar irrigation and growing systems, harvest the
grapes at the same stage of ripeness, and so on, their wines should
be very similar— very similar, that is, but also very different according
to the winemakers because every tiny difference in the land and meth-
ods is magnified in the final product. Sometimes the claimed differ-
ences are so small as to be silly: In Bordeaux, a wine-producing re-
gion in France, a dirt road can separate two vineyards, yet the wine
from one may cost several times that of the other.
When performing a statistical analysis of two populations, think
about crossing that dirt road and how that tiny difference might cloud
your results.
Glossary
d The difference between two matched values in two samples col-
lected from the same source. It is called the paired difference.
d The mean of the paired differences for a sample.
Independent samples Two samples drawn from two populations
such that the selection of one does not affect the selection of the other.
Paired or matched samples Two samples drawn in such a way
that they include the same elements and two data values are ob-
tained from each element, one for each sample. Also called depend-
ent samples.
fi d The mean of the paired differences for the population.
s d The standard deviation of the paired differences for a sample.
<T d The standard deviation of the paired differences for the
population.
Supplementary Exercises
10.76 A consulting agency was asked by a large insurance company to investigate if business majors were
better salespersons than those with other majors. A sample of 20 salespersons with a business degree
showed that they sold an average of 1 1 insurance policies per week. Another sample of 25 salespersons
with a degree other than business showed that they sold an average of 9 insurance policies per week.
484 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Assume that the two populations are normally distributed with population standard deviations of 1.80 and
1.35 policies per week, respectively.
a. Construct a 99% confidence interval for the difference between the two population means.
b. Using the 1% significance level, can you conclude that persons with a business degree are bet-
ter salespersons than those who have a degree in another area?
10.77 According to an estimate, the average earnings of female workers who are not union members are
$388 per week and those of female workers who are union members are $505 per week. Suppose that
these average earnings are calculated based on random samples of 1500 female workers who are not union
members and 2000 female workers who are union members. Further assume that the standard deviations
for the two corresponding populations are $30 and $35, respectively.
a. Construct a 95% confidence interval for the difference between the two population means.
b. Test at the 2.5% significance level whether the mean weekly earnings of female workers who
are not union members are less than those of female workers who are union members.
10.78 An economist was interested in studying the impact of the recession on dining out, including
drive-thru meals at fast food restaurants. A random sample of forty-eight families of four with discre-
tionary incomes between $300 and $400 per week indicated that they reduced their spending on dining
out by an average of $31.47 per week, with a sample standard deviation of $10.95. Another random
sample of 42 families of five with discretionary incomes between $300 and $400 per week reduced their
spending on dining out by an average $35.28 per week, with a sample standard deviation of $12.37.
(Note that the two groups of families are differentiated by the number of family members.) Assume that
the distributions of reductions in weekly dining-out spendings for the two groups have the same popu-
lation standard deviation.
a. Construct a 90% confidence interval for the difference in the mean weekly reduction in dining-
out spending levels for the two populations.
b. Using the 5% significance level, can you conclude that the average weekly spending reduction
for all families of four with discretionary incomes between $300 and $400 per week is less than
the average weekly spending reduction for all families of five with discretionary incomes be-
tween $300 and $400 per week?
10.79 In 2007, the average number of fatalities in all railroad accidents was .1994 per accident, whereas
the average number of fatalities in all recreational boating accidents was .1320 per accident (Source:
http://www.bts.gOv/publications/national_transportation_statistics/#chapter_4). Suppose that random sam-
ples of railroad and recreational boating accidents for this year have average numbers of fatalities of .183
and .146 per accident, with standard deviations of .82 and .67, respectively. The railroad statistics are based
on a random sample of 418 accidents, whereas the boating statistics are based on a random sample of 392
accidents. Assume that the distributions of the numbers of fatalities have the same population standard de-
viation for the two groups.
a. Construct a 98% confidence interval for the difference in the average number of fatalities per
accident in all railroad and all recreational boating accidents.
b. Using the 1% significance level, can you conclude that the average number of fatalities in all
railroad accidents is higher than the average number of fatalities in all recreational boating ac-
cidents?
10.80 The manager of a factory has devised a detailed plan for evacuating the building as quickly as
possible in the event of a fire or other emergency. An industrial psychologist believes that workers ac-
tually leave the factory faster at closing time without following any system. The company holds fire
drills periodically in which a bell sounds and workers leave the building according to the system. The
evacuation time for each drill is recorded. For comparison, the psychologist also records the evacua-
tion time when the bell sounds for closing time each day. A random sample of 36 fire drills showed a
mean evacuation time of 5.1 minutes with a standard deviation of 1.1 minutes. A random sample of
37 days at closing time showed a mean evacuation time of 4.2 minutes with a standard deviation of
1.0 minute.
a. Construct a 99% confidence interval for the difference between the two population means.
b. Test at the 5% significance level whether the mean evacuation time is smaller at closing time
than during fire drills.
Assume that the evacuation times at closing time and during fire drills have equal but unknown popula-
tion standard deviations.
10.81 Two local post offices are interested in knowing the average number of Christmas cards that are
mailed out from the towns that they serve. A random sample of 80 households from Town A showed that
they mailed an average of 28.55 Christmas cards with a standard deviation of 10.30. The corresponding
values of the mean and standard deviation produced by a random sample of 58 households from Town B
Supplementary Exercises 485
were 33.67 and 8.97 Christmas cards. Assume that the distributions of the numbers of Christmas cards
mailed by all households from both these towns have the same population standard deviation.
a. Construct a 95% confidence interval for the difference in the average numbers of Christmas cards
mailed by all households in these two towns.
b. Using the 10% significance level, can you conclude that the average number of Christmas cards
mailed out by all households in Town A is different from the corresponding average for Town B?
10.82 Refer to Exercise 10.78. Now answer the questions of parts a and b there without assuming that the
standard deviations are the same for the two populations but under the following two situations.
a. Using the sample standard deviations given in Exercise 10.78.
b. Using sample standard deviations of $7.17 and $15.80 for families of four and families of five,
respectively.
10.83 Repeat Exercise 10.79 without assuming that the standard deviations of the two populations are the
same but considering the following two situations.
a. Using the sample standard deviations given in Exercise 10.79.
b. Using sample standard deviations of .91 and .39 for railroad and recreational boating accidents,
respectively.
10.84 Repeat Exercise 10.80 without assuming that the standard deviations for the two populations are
the same but considering the following two situations.
a. Using the sample standard deviations given in Exercise 10.80.
b. Using sample standard deviations of 1.33 and .72 for fire drills and closing time, respectively.
10.85 Repeat Exercise 10.81 without assuming that the standard deviations for the two populations are
the same but considering the following two situations.
a. Using the sample standard deviations given in Exercise 10.81.
b. Using sample standard deviations of 6.85 and 11.97 for Town A and Town B, respectively.
10.86 The owner of a mosquito-infested fishing camp in Alaska wants to test the effectiveness of two ri-
val brands of mosquito repellents, X and Y. During the first month of the season, eight people are chosen
at random from those guests who agree to take part in the experiment. For each of these guests, Brand X
is randomly applied to one arm and Brand Y is applied to the other arm. These guests fish for 4 hours,
then the owner counts the number of bites on each arm. The table below shows the number of bites on
the arm with Brand X and those on the arm with Brand Y for each guest.
Guest
A
B
C
D
E
F
G
H
Brand X
12
23
18
36
8
27
22
32
Brand Y
9
20
21
27
6
18
15
25
a. Construct a 95% confidence interval for the mean [L d of population paired differences, where a
paired difference is defined as the number of bites on the arm with Brand X minus the number
of bites on the arm with Brand Y.
b. Test at the 5% significance level whether the mean number of bites on the arm with Brand X
and the mean number of bites on the arm with Brand Y are different for all such guests.
Assume that the population of paired differences has a normal distribution.
10.87 A random sample of nine students was selected to test for the effectiveness of a special course
designed to improve memory. The following table gives the scores in a memory test given to these stu-
dents before and after this course.
Before
43
57
48
65
81
49
38
69
58
After
49
56
55
77
89
57
36
64
69
a. Construct a 95% confidence interval for the mean \x, d of the population paired differences, where
a paired difference is defined as the difference between the memory test scores of a student be-
fore and after attending this course.
b. Test at the 1% significance level whether this course makes any statistically significant
improvement in the memory of all students.
Assume that the population of the paired differences has a normal distribution.
10.88 In a random sample of 800 men aged 25 to 35 years, 24% said they live with one or both parents.
In another sample of 850 women of the same age group, 18% said that they live with one or both parents.
486 Chapter 10 Estimation and Hypothesis Testing: Two Populations
a. Construct a 95% confidence interval for the difference between the proportions of all men and
all women aged 25 to 35 years who live with one or both parents.
b. Test at the 2% significance level whether the two population proportions are different.
c. Repeat the test of part b using the p-value approach.
10.89 In a Pew Research survey from July 2009, participants were asked about the economic news that
they had heard since May 2009. Forty-five percent of Independents stated that the news was mostly bad,
as opposed to 48% of Republicans and 30% of Democrats. The study mentioned a total sample size of
1001. Assume that this sample included 395, 286, and 320 Independents, Republicans, and Democrats,
respectively.
a. Construct a 95% confidence interval for the difference in the population proportions for each of
the three pairs of political affiliations.
b. At the 1% level, can you conclude that the percentage of Independents who stated that the news
was mostly bad was significantly different from the corresponding percentage of Republicans?
Use the critical value method.
c Repeat part b using the p- value method.
10.90 A June 2009 Harris Interactive poll asked people their opinions about the influence of advertising
on the products they buy. Among the people aged 18 to 34 years, 45% view advertisements as being in-
fluential, whereas among the people aged 35 to 44 years, 37% view advertisements as being influential.
Suppose that this survey included 655 people in the 18- to 34-year age group and 420 in the 35- to 44-
year age group.
a. Find a 98% confidence interval for the difference in the population proportions for the two age
groups.
b. At the 1% significance level, can you conclude that the proportion of all people aged 18 to 34
years who view advertisements as being influential in their purchases is greater than the propor-
tion of all people aged 35 to 44 years who hold the same opinion?
10.91 A June 2009 Gallup Poll asked a sample of Americans whether they trust specific groups or indi-
viduals when it comes to making recommendations about healthcare reform. Sixty percent of Democrats
and 68% of Republicans stated that they trust doctors' opinions about healthcare reform (Source:
http://www.gallup.com/poll/120890/Healthcare-Americans-Trust-Physicians-Politicians.aspx). Suppose this
survey included 340 Democrats and 306 Republicans.
a. Make a 90% confidence interval for the difference in the population proportions for the two
groups of people.
b. At the 5% significance level, can you conclude that the proportion of all Democrats who trust
doctors' opinions about healthcare reform differs from the proportion of all Republicans who
trust doctors' opinions about healthcare reform?
Advanced Exercises
10.92 Manufacturers of two competing automobile models, Gofer and Diplomat, each claim to have
the lowest mean fuel consumption. Let p t be the mean fuel consumption in miles per gallon (mpg) for
the Gofer and p 2 me mean fuel consumption in mpg for the Diplomat. The two manufacturers have
agreed to a test in which several cars of each model will be driven on a 100-mile test run. Then the fuel
consumption, in mpg, will be calculated for each test run. The average of the mpg for all 100-mile test
runs for each model gives the corresponding mean. Assume that for each model the gas mileages for
the test runs are normally distributed with a = 2 mpg. Note that each car is driven for one and only
one 100-mile test run.
a. How many cars (i.e., sample size) for each model are required to estimate p x — p 2 w i m a 90%
confidence level and with a margin of error of estimate of 1.5 mpg? Use the same number of
cars (i.e., sample size) for each model.
b. If p { is actually 33 mpg and p 2 ls actually 30 mpg, what is the probability that five cars for each
model would yield x x S x 2 ?
10.93 Maria and Ellen both specialize in throwing the javelin. Maria throws the javelin a mean distance
of 200 feet with a standard deviation of 10 feet, whereas Ellen throws the javelin a mean distance of
210 feet with a standard deviation of 12 feet. Assume that the distances each of these athletes throws
the javelin are normally distributed with these population means and standard deviations. If Maria and
Ellen each throw the javelin once, what is the probability that Maria's throw is longer than Ellen's?
10.94 A new type of sleeping pill is tested against an older, standard pill. Two thousand insomniacs are
randomly divided into two equal groups. The first group is given the old pill, and the second group receives
the new pill. The time required to fall asleep after the pill is administered is recorded for each person. The
Supplementary Exercises 487
results of the experiment are given in the following table, where x and s represent the mean and standard
deviation, respectively, for the times required to fall asleep for people in each group after the pill is taken.
Group 1
Group 2
(Old Pill)
(New Pill)
II
1000
1000
X
15.4 minutes
15.0 minutes
s
3.5 minutes
3.0 minutes
Consider the test of hypothesis H ; (jl x — p 2 = versus H x : ix x — /jl 2 > 0, where (jl x and p 2 are the mean
times required for all potential users to fall asleep using the old pill and the new pill, respectively.
a. Find the p-value for this test.
b. Does your answer to part a indicate that the result is statistically significant? Use a = .025.
c. Find the 95% confidence interval for p x — /jl 2 .
d. Does your answer to part c imply that this result is of great practical significance?
10.95 Gamma Corporation is considering the installation of governors on cars driven by its sales staff. These
devices would limit the car speeds to a preset level, which is expected to improve fuel economy. The com-
pany is planning to test several cars for fuel consumption without governors for 1 week. Then governors
would be installed in the same cars, and fuel consumption will be monitored for another week. Gamma Cor-
poration wants to estimate the mean difference in fuel consumption with a margin of error of estimate of 2
mpg with a 90% confidence level. Assume that the differences in fuel consumption are normally distributed
and that previous studies suggest that an estimate of s d = 3 mpg is reasonable. How many cars should be
tested? (Note that the critical value of t will depend on n, so it will be necessary to use trial and error.)
10.96 Refer to Exercise 10.95. Suppose Gamma Corporation decides to test governors on seven cars. However,
the management is afraid that the speed limit imposed by the governors will reduce the number of contacts the
salespersons can make each day. Thus, both the fuel consumption and the number of contacts made are recorded
for each car/salesperson for each week of the testing period, both before and after the installation of governors.
Salesperson
Number of Contacts
Fuel Consumption (mpg)
Before
After
Before
After
A
50
49
25
26
B
63
60
21
24
C
42
47
27
26
D
55
51
23
25
E
44
50
19
24
F
65
60
18
22
G
66
58
20
23
Suppose that as a statistical analyst with the company, you are directed to prepare a brief report that
includes statistical analysis and interpretation of the data. Management will use your report to help
decide whether or not to install governors on all salespersons' cars. Use 90% confidence intervals and
.05 significance levels for any hypothesis tests to make suggestions. Assume that the differences in
fuel consumption and the differences in the number of contacts are both normally distributed.
10.97 Two competing airlines, Alpha and Beta, fly a route between Des Moines, Iowa, and Wichita,
Kansas. Each airline claims to have a lower percentage of flights that arrive late. Let p l be the propor-
tion of Alpha's flights that arrive late and p 2 the proportion of Beta's flights that arrive late.
a. You are asked to observe a random sample of arrivals for each airline to estimate p l — p 2 with a
90% confidence level and a margin of error of estimate of .05. How many arrivals for each airline
would you have to observe? (Assume that you will observe the same number of arrivals, n, for each
airline. To be sure of taking a large enough sample, use p x = p 2 = .50 in your calculations for n.)
b. Suppose that p x is actually .30 and p 2 is actually .23. What is the probability that a sample of
100 flights for each airline (200 in all) would yield p x a p 2 l
10.98 Refer to Exercise 10.56, in which a random sample of six cars was selected to test a gasoline addi-
tive. The six cars were driven for 1 week without the gasoline additive and then for 1 week with the addi-
tive. The data reproduced here from that exercise show miles per gallon without and with the additive.
488 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Without
24.6
28.3
18.9
23.7
15.4
29.5
With
26.3
31.7
18.2
25.3
18.3
30.9
Suppose that instead of the study with 6 cars, a random sample of 12 cars is selected and these cars are
divided randomly into two groups of 6 cars each. The cars in the first group are driven for 1 week with-
out the additive, and the cars in the second group are driven for 1 week with the additive. Suppose that
the top row of the table lists the gas mileages for the 6 cars without the additive, and the bottom row
gives the gas mileages for the cars with the additive. Assume that the distributions of the gas mileages
with or without the additive are (approximately) normal with equal but unknown standard deviations.
a. Would a paired sample test as described in Section 10.4 be appropriate in this case? Why or
why not? Explain.
b. If the paired sample test is inappropriate here, carry out a suitable test of whether the mean gas
mileage is lower without the additive. Use a = .025.
c. Compare your conclusion in part b with the result of the hypothesis test in Exercise 10.56.
10.99 Does the use of cellular telephones increase the risk of brain tumors? Suppose that a manufacturer
of cell phones hires you to answer this question because of concern about public liability suits. How would
you conduct an experiment to address this question? Be specific. Explain how you would observe, how
many observations you would take, and how you would analyze the data once you collect them. What are
your null and alternative hypotheses? Would you want to use a higher or a lower significance level for the
test? Explain.
10.100 We wish to estimate the difference between the mean scores on a standardized test of students
taught by Instructors A and B. The scores of all students taught by Instructor A have a normal distri-
bution with a standard deviation of 15, and the scores of all students taught by Instructor B have a nor-
mal distribution with a standard deviation of 10. To estimate the difference between the two means,
you decide that the same number of students from each instructor's class should be observed.
a. Assuming that the sample size is the same for each instructor's class, how large a sample should
be taken from each class to estimate the difference between the mean scores of the two popu-
lations to within 5 points with 90% confidence?
b. Suppose that samples of the size computed in part a will be selected in order to test for the dif-
ference between the two population mean scores using a .05 level of significance. How large
does the difference between the two sample means have to be for you to conclude that the two
population means are different?
c. Explain why a paired-samples design would be inappropriate for comparing the scores of In-
structor A versus Instructor B.
10.101 The weekly weight losses of all dieters on Diet I have a normal distribution with a mean of 1.3
pounds and a standard deviation of .4 pound. The weekly weight losses of all dieters on Diet II have a
normal distribution with a mean of 1.5 pounds and a standard deviation of .7 pound. A random sample of
25 dieters on Diet I and another sample of 36 dieters on Diet II are observed.
a. What is the probability that the difference between the two sample means, x x — x 2 , will be
within -.15 to .15, that is, -.15 < x, - x 2 < .15?
b. What is the probability that the average weight loss x l for dieters on Diet I will be greater than
the average weight loss x 2 for dieters on Diet II?
c. If the average weight loss of the 25 dieters using Diet I is computed to be 2.0 pounds, what is
the probability that the difference between the two sample means, x l — x 2 , will be within —.15
to .15, that is, -.15 < x, - x 2 < .15?
d. Suppose you conclude that the assumption —.15 < — /jl 2 < .15 is reasonable. What does
this mean to a person who chooses one of these diets?
10.102 Sixty-five percent of all male voters and 40% of all female voters favor a particular candidate. A
sample of 100 male voters and another sample of 100 female voters will be polled. What is the probabil-
ity that at least 10 more male voters than female voters will favor this candidate?
Self-Review Test
1. To test the hypothesis that the mean blood pressure of university professors is lower than that of com-
pany executives, which of the following would you use?
a. A left-tailed test b. A two-tailed test c. A right-tailed test
Mini-Projects 489
2. Briefly explain the meaning of independent and dependent samples. Give one example of each of
these cases.
3. A company psychologist wanted to test if company executives have job-related stress scores higher
than those of university professors. He took a sample of 40 executives and 50 professors and tested them
for job-related stress. The sample of 40 executives gave a mean stress score of 7.6. The sample of 50 pro-
fessors produced a mean stress score of 5.4. Assume that the standard deviations of the two populations
are .8 and 1.3, respectively.
a. Construct a 99% confidence interval for the difference between the mean stress scores of all
executives and all professors.
b. Test at the 2.5% significance level whether the mean stress score of all executives is higher than
that of all professors.
4. A sample of 20 alcoholic fathers showed that they spend an average of 2.3 hours per week playing
with their children with a standard deviation of .54 hour. A sample of 25 nonalcoholic fathers gave a mean
of 4.6 hours per week with a standard deviation of .8 hour.
a. Construct a 95% confidence interval for the difference between the mean times spent per week
playing with their children by all alcoholic and all nonalcoholic fathers.
b. Test at the 1% significance level whether the mean time spent per week playing with their
children by all alcoholic fathers is less than that of nonalcoholic fathers.
Assume that the times spent per week playing with their children by all alcoholic and all nonalcoholic
fathers both are normally distributed with equal but unknown standard deviations.
5. Repeat Problem 4 assuming that the times spent per week playing with their children by all alco-
holic and all nonalcoholic fathers both are normally distributed with unequal and unknown standard
deviations.
6. Lake City has two shops, Zeke's and Elmer's, that handle the majority of the town's auto body re-
pairs. Seven cars that were damaged in collisions were taken to both shops for written estimates of the re-
pair costs. These estimates (in dollars) are shown in the following table.
Zeke's
1058
544
1349
1296
676
998
1698
Elmer's
995
540
1175
1350
605
970
1520
a. Construct a 99% confidence interval for the mean \l a of the population paired differences, where
a paired difference is equal to Zeke's estimate minus Elmer's estimate.
b. Test at the 5% significance level whether the mean fi d of the population paired differences is dif-
ferent from zero.
Assume that the population of paired differences is (approximately) normally distributed.
7. A sample of 500 male registered voters showed that 57% of them voted in the last presidential
election. Another sample of 400 female registered voters showed that 55% of them voted in the same
election.
a. Construct a 97% confidence interval for the difference between the proportions of all male and all
female registered voters who voted in the last presidential election.
b. Test at the 1% significance level whether the proportion of all male voters who voted in the last
presidential election is different from that of all female voters.
Mini-Projects
■ MINI-PROJECT 10-1
Suppose that a new cold-prevention drug was tested in a randomized, placebo-controlled, double-blind
experiment during the month of January. One thousand healthy adults were randomly divided into two
groups of 500 each — a treatment group and a control group. The treatment group was given the new drug,
and the control group received a placebo. During the month, 40 people in the treatment group and 120
people in the control group caught a cold. Explain how to construct a 95% confidence interval for the dif-
ference between the relevant population proportions. Also describe an appropriate hypothesis test, using
the given data, to evaluate the effectiveness of this new drug for cold prevention.
Find a similar article in a journal of medicine, psychology, or other field that lends itself to confi-
dence intervals and hypothesis tests for differences in two means or proportions. First explain how to make
the confidence intervals and hypothesis tests; then do so using the data given in the article.
490 Chapter 10 Estimation and Hypothesis Testing: Two Populations
■ MINI-PROJECT 10-2
A researcher conjectures that cities in the more populous states of the United States tend to have higher
costs for doctors' visits. Using "CITY DATA" that accompany this text, select a random sample of 10
cities from the six most populous states (California, Texas, New York, Florida, Illinois and Pennsylvania).
Then take a random sample of 10 cities from the remaining states in the data set. For each of the 20 cities,
record the average cost of a doctor's visit. Assume that such costs are approximately normally distributed
for all cities in each of the two groups of states. Further assume that the cities you selected make random
samples of all cities for the two groups of states. Assume that the standard deviations for the two groups
are unequal and unknown.
a. Construct a 95% confidence interval for the difference in the mean costs of doctors' visits for all
cities in the two groups of states.
b. At the 5% level of significance, can you conclude that the average cost of a doctor's visit for all
cities in the six most populous states is higher than that of a doctor's visit for all cities in the re-
maining states?
■ MINI-PROJECT 10-3
Many different kinds of analyses have been performed on the salaries of professional athletes. Perform
a hypothesis test of whether or not the average salaries of players in two sports are different by tak-
ing independent random samples of 35 players each from any two sports of your choice from Major
League Baseball (MLB), the National Football League (NFL), the National Basketball Association
(NBA), and the National Hockey League (NHL). (Note: A good Internet reference for such data is
http://www.usatoday.com/sports/salaries/index.htm.) After you take samples, do the following.
a. For each player, calculate the weekly salary. For your information, the approximate length (in
weeks) of a season is 32.5 for MLB, 22.5 for the NFL, 28 for the NBA, and 29.5 for the NHL.
This length of a season does not include the playoffs, but it does include training camp and the
preseason games because each player is expected to participate in these events. Players may re-
ceive bonuses for making the playoffs, but these are not included in their base salaries. You may
ignore such bonuses.
b. Perform a hypothesis test to determine if the average weekly salaries are the same for the two
sports that you selected. Use a significance level of 5%. Make certain to indicate whether you de-
cide to use the pooled variance assumption or not, and justify your selection.
c. Perform a hypothesis test on the same data to determine if the average annual salaries are the same
for the two sports that you selected. Explain why you could get a different answer (with regard to
rejecting or failing to reject the null hypothesis) when using the weekly salaries versus the annual
salaries.
■ MINI-PROJECT 10-4
Refer to Case Study 9-2 that discussed the results of a survey in which adults were asked what seat they
prefer on a plane when they fly: window, middle, or aisle seat. Obtain random samples of 60 male and 60
female college students and ask the following question: When you fly on a plane, do you prefer to have
a window seat or a non-window seat? Perform a hypothesis test to determine if the proportion of female
college students who prefer to have a window seat when flying is different from the proportion of male
college students who prefer to have a window seat when flying. Use a 5% significance level.
■ MINI-PROJECT 10-5
Refer to Case Study 8-1, which discussed the results of a USDA study about the cost of raising a child.
As was noted in that Case Study, the average expenditure to raise a child born in 2008 through the age of
17 for families in each of the three income levels were as shown in the following table.
Family Income
Mean Expenditure
Less than $56,870
$159,870
$56,870 to $98,470
$221,190
More than $98,470
$366,660
a. Suppose that the equal variance assumption is reasonable for the lower and middle income groups
(the first two groups listed in the table). Using the "Less than $56,870" group as population 1 and
the "$56,870 to $98,470" group as population 2, determine the largest possible value of the pooled
Technology Instruction 491
standard deviation that would cause one to reject the null hypothesis H : p, x = p 2 in favor of the
hyphothesis H x : p x < /x 2 at the 5% significance level when n t = n 2 =10.
b. Repeat part a for the following sample sizes
i. n x = n 2 = 15
ii. n x = n 2 = 20
iii. n l = n 2 = 30
c. The sample size for three groups (combined) in the study was more than 3000. Based on your re-
sults in parts a and b, do you think that it would have been possible to obtain a pooled standard
deviation that would have caused you not to reject H a : pu X = p 2 in favor of the hypothesis
H x : < ijl 2 at the 5% significance level? Explain why.
DECIDE FOR YOURSELF
Deciding About How to Design a Study
By now, you might feel that you have learned almost everything there
is to know about statistics. In some ways, you have learned a great
deal. When using the p- value approach, the rule to reject a null hypoth-
esis whenever the p-value is less than the significance level never
changes. If you know this rule, you do not have to worry about chang-
ing it. You have also learned the basic concept of a confidence inter-
val, which will also never change. However, one of the most important
lessons to learn in statistics is how to conduct a valid study. Design of
experiments and sampling design are two areas of statistics that are
dedicated to determining the proper way to plan a study before any
data are collected. Without a proper plan, the time and money spent on
the study could be a complete waste if the results are not valid.
Consider the example of gasoline additive mentioned in the Decide
for Yourself section of Chapter 9. In that section, we discussed perform-
ing a single-sample procedure. However, the same problem could be
addressed by using some of the procedures learned in this chapter.
1. Describe how that analysis could be performed by selecting two
independent samples of cars. Be specific about how the treatments
are applied/assigned to the cars, whether there are any special con-
siderations as to how the cars are selected, and the specific measure-
ments that would be compared.
2. Answer question 1 assuming that we use a paired-sample proce-
dure instead of a two independent samples procedure.
3. Discuss the strengths and weaknesses of the three procedures
(including the single sample procedure discussed in Chapter 9).
Which method would you prefer and why? Explain.
ECHNOLOGY
INSTRUCTION
Confidence Intervals and Hypothesis Tests for Two Populations
1
2-SanpTTest
ListlVLi
List2:l_2
Fre-=il : IB
Stats
Fre-=i2: 1
JPooJ
<v-2 >v-2
'•m Yes
Screen 10.1
2-SanpTTest
t= -1.204796141
p=. 2522070299
df= 11.63973844
xi=22.375
ix2=24
Screen 10.2
1. To perform a hypothesis test about the difference between the means of two populations
with independent samples, select STAT >TESTS >2-SampTtest. If the data are stored
in lists, select Data, and enter the names of the lists. If, instead, you have summary statis-
tics for the two samples, select Stats, and enter the mean, standard deviation, and sample
size for each sample. Choose the form of the alternative hypothesis. If you are assuming
that the standard deviations are equal for the two populations, select Yes for Pooled;
otherwise, select No. Select Calculate to find the p-value. (See Screens 10.1 and 10.2.)
2. To perform a hypothesis test about the proportions of two populations using independent
samples, select STAT >TESTS >2-PropZTest. Enter the successes and trials (as x and n
respectively) for each of the two samples. Select the alternative hypothesis, and then
Calculate to find the /j-value of the test. Be careful to distinguish between the /7-value
and the sample proportions, which have hats above them.
3. To find a confidence interval for the difference of the means of two populations using in-
dependent samples, select STAT >TESTS >2-SampTInt. If the data is stored in lists,
select Data, and enter the names of the lists. If, instead, you have summary statistics for
the two samples, select Stats, and enter the mean, standard deviation, and sample size for
each sample. Enter the confidence level as the C-Level. If you are assuming that the
standard deviations are equal for the two populations, select Yes for Pooled; otherwise,
select No. Select Calculate to find the confidence interval.
492 Chapter 10 Estimation and Hypothesis Testing: Two Populations
4. To find a confidence interval for the difference between two population proportions, select
STATS >TESTS >2-PropZInt. Enter the successes and trials (as x and n, respectively) for
each of the two samples. Enter the confidence level, and then select Calculate to find the
confidence interval.
1. To find a confidence interval for fi 1 — fi 2 for two independent populations with unknown
but equal standard deviations as discussed in Section 10.2, select Stat >Basic Statistics
>2-Sample t. In the dialog box you obtain, select Summarized data, and enter the val-
ues of the Sample sizes, (sample) Means, and Standard deviations for the two samples.
Check the box next to Assume equal variances. Click the Options button, and enter the
value of the Confidence level in the new dialog box. Click OK in both boxes. The output
containing the confidence interval will appear in the session window.
If instead of summary measures, you have data on the two samples, enter these data
in columns CI and C2 of the Minitab spreadsheet. In the dialog box, click next to Sam-
ples in different columns, and enter the column names for two samples. (See Screens
10.3 and 10.4.) The rest of the procedure is the same as above.
2. To perform a hypothesis test about — jx 2 for two independent populations with unknown
but equal standard deviations as discussed in Section 10.2, select Stat >Basic Statistics >
2-Sample t. In the dialog box you obtain, select Summarized data, and enter the values of
the Sample sizes, (sample) Means, and Standard deviations for the two samples. Check the
box next to Assume equal variances. Click the Options button. In the new dialog box you
obtain, enter for the Test difference, and select the appropriate Alternative hypothesis.
Click OK in both boxes. The output containing the p-value will appear in the Session window.
If, instead of summary measures, you have data on the two samples, enter these data
in columns CI and C2 of the Minitab spreadsheet. In the dialog box, click next to Sam-
ples in different columns, and enter the column names for two samples. The rest of the
procedure is the same as above.
2 Sample t (Test and Confidence Interval)
CI
C2
Sample 1
Sample 2
C Samples in one column
Samples! |
Subscripts: f
<• Samples in different columns
First: |' Sample 1'
Second: | 1 Sample 2 '
f" Summarized data
Sample size:
First: |
Second: f
r Assume equal variances
Mean:
Standard
deviation:
Selecl
Graphs...
Help
OK
Options...
Cancel
Screen 10.3
To find a confidence interval for i± x — fju 2
or to perform a hypothesis test about fi l —
jjL 2 for two independent populations with un-
known and unequal standard deviations
discussed in Section 10.3, the procedures
are the same as in steps 1 and 2 above, re-
spectively, except that you do not check
next to Assume equal variances.
To find a confidence interval for ix d for
paired data discussed in Section 10.4, enter
the Before and After data into columns CI
and C2, respectively. Select Stat >Basic
Statistics >Paired t. In the dialog box
you obtain, select Samples in columns,
and enter the column names CI and C2 in
the boxes next to First sample and Sec-
ond sample. Click the Options button, and
enter the value of the Confidence level in
the new dialog box. Click OK in both
boxes. The output containing the confidence
interval will appear in the session window. Note that the confidence interval here is for the
mean of the differences given by CI — C2, which represents Before — After.
Technology Instruction 493
Two-Sample T-Test and CI: Sample 1 , Sample 2
Two-sample T for Sample 1 vs Sample 2
H Mean StDev SE Mean
Sample 1 13 29.62 4.99 1.4
Sample 2 12 25.50 3.23 0.93
Difference = mu [Sample 1) - mu [Sample 2)
Estimate for difference: 4.11538
95% CI for difference: [0.63216, 7.59861)
T-Test of difference = [vs not =.) : T-Value
2.46 P-Value = 0.023 DF
20
Screen 10.4
Paired t (Test and Confidence Interval)
<" Samples in columns
First sample: |
Second sample: [~
* Summarized data (differences]
Sample size: [7
Mean: fs
Standard deviation: jlO . 7858
Paired t evaluates the first sample
minus the second sample.
Select
Graphs...
Options.
Help
OK
Cancel
Screen 10.5
Paired T-Test and CI
Difference 7
Mean
5.00000
StDev
10.78580
95% CI for mean difference:
T-Test of mean difference =
Screen 10.6
To perform a hypothesis test about fji d for paired data
discussed in Section 10.4, enter the Before and After data
into columns CI and C2, respectively. Select Stat >Basic
Statistics >Paired t. In the dialog box you obtain, select
Samples in columns, and enter the column names CI
and C2 in the boxes next to First sample and Second
sample. Click the Options button. In the new dialog box
you obtain, enter for the Test mean, and select the
appropriate Alternative hypothesis. Click OK in both
boxes. The output containing the p-value will appear in
the session window. Note that the hypothesis test here is
for the mean of the differences given by CI — C2, which
represents Before — After. You need to keep this in mind
when determining your alternative hypothesis. (See
Screens 10.5 and 10.6.)
To find a confidence interval for p x — p 2 using two large
and independent samples as discussed in Section 10.5,
select Stat >Basic Statistics >2 Proportions. In the
dialog box you obtain, click on Summarized data, and
enter the sample sizes and the numbers of
successes in the boxes below Trials and
Events, respectively, for the two samples.
Click the Options button, and enter the value
of the Confidence Level in the new dialog
box. Click OK in both dialog boxes. The
output containing the confidence interval for
P\ ~ Pi w iU appear in the session window.
To perform a hypothesis test about p x — p 2 using two large and independent samples as
discussed in Section 10.5, select Stat >Basic Statistics >2 Proportions. In the dialog
box you obtain, select Summarized data, and then enter the sample sizes and the num-
bers of successes in the boxes below Trials and Events, respectively, for the two samples.
Click the Options button. Set Test difference to 0, select the appropriate Alternative hy-
pothesis, and check next to Use pooled estimate of p for test in the new dialog box.
Click OK in both dialog boxes. The output containing the p-value for the test will appear
in the session window. (See Screens 10.7 and 10.8.)
SE Mean
4.07665
(-4.97520, 14.97520)
(vs not = 0) : T-Value
1.23 P-Value = 0.266
4 94 Chapter 10 Estimation and
Hypothesis Testing: Two Populations
Screen 10.7
2 Proportions (Test and Confidence Interval)
r Samples in one column
Samples:
Subscripts: |~
C Samples in different columns
First:
Second:
'* Summarized data
Trials:
First:
Second:
500
Events:
[Too
400
63
Select
Options..
Help
OK
Cancel
Screen 10.8
Test and CI for Two Proportions
Sample X
1 100
2 68
N Sample p
500 0.200000
400 0.170000
Difference = p [1) - p [2)
Estimate for difference: 0.03
95k CI for difference: [-0.0208364, 0.0808364)
Test for difference = (vs not =0): 2 = 1.16 P-Value
0.247
The Data Analysis ToolPak contains preprogrammed functions for performing the following tests
• The paired f-test
• The two-independent-sample f-test for means, assuming equal variances
• The two-independent-sample f-test for means, assuming unequal variances
The dialog boxes for all three tests are set up in exactly the same fashion. Hence, no matter which
test you are using, the processes of entering the data ranges, the hypothesized difference, and so
on, are the same for all three tests. Although there is no restriction on the location of the data in
the spreadsheet, the instructions will be provided assuming that the data are in adjacent columns.
1. Click the Data tab. Click the Data Analysis button within the Analysis group. From the
Data Analysis window that appears, select the appropriate test from the list:
• t-test: Paired Two Sample for Means
• t-test: Two-Sample Assuming Equal Variances
• t-test: Two-Sample Assuming Unequal Variances
2. Enter the location of first set of paired data in the Variable 1 Range box. Enter the
location of the second set of paired data in the Variable 2 Range box. Excel will always
create differences in the order "variable 1 - variable 2." Enter the value for the
hypothesized difference from the null hypothesis in the Hypothesized Mean Difference
box. Enter the significance level, as a decimal, in the Alpha box. If your columns of data
have labels in the top row, click the Labels box. Choose how you wish for the output to
appear. (See Screen 10.9.) Click OK.
Technology Assignments 495
t-Test: Paired Two Sample for Means ? X
Screen 10.9
A
4
1
\ K i'i4L-
witn
2
26.3
24.6
a
Jl. /
28.3
4
lB.Z
la.y
5
2b. J
23./
D
18. a
10.4
—r
I
30.9
/y.o
8
10
11
12
13
14
Input
Variable I Range:
Variable 2 Range:
Hypothesized Mean Difference:
0yjbeis|
Alpha: 1 0.05
$A$1:$A$7
$B$lr.$B$7
OK
Cancel
Help
Output options
O Output Range:
© New Worksheet Ply:
O New Workbook
A
B
C
1
t-Test: Paired Two Sample for Means
2
With
Without
4
Mean
25.11667
23.4
5
Variance
34.50567
29.4
6
Observations
6
6
7
Pearson Correlation
0.371219
8
Hypothesized Mean Difference
9
df
5
10
t Stat
2.945744
11
P(T<=t) one-tail
0.016021
12
t Critical one-tail
2.015O4S
13
P(T<=t) two-tail
0.032043
14
t Critical two -tail
2.570582
3. The two lines in the output that you will need to deter-
mine the /7-value are the lines labeled t Stat and
P(T<=t) two-tail. (See Screen 10.10.) If the alterna-
tive hypothesis is two-tailed, the value in the P(T<=t)
two-tail box is the />value for the test. If the alternative
hypothesis is one-tailed, use the following set of rules:
a. If the hypothesis test is left-tailed and the value of
t Stat is negative OR the hypothesis test is right-
tailed and the value of t Stat is positive, the p-value
of the test is equal to one-half the value in the
P(T<=t) two-tail box.
b. If the hypothesis test is left-tailed and the value of
t Stat is positive OR the hypothesis test is right-
tailed and the value of t Stat is negative, the p-
value of the test is equal to 1 minus one-half the
value in the P(T<=t) two-tail box.
Note: Screen 10.10 shows the output for a paired f-test.
The output windows for the independent samples tests are
very similar. More important, the instructions given in
step 3 hold for all three types of tests.
Screen 10.10
TECHNOLOGY ASSIGNMENTS
TA10.1 Fifty randomly selected 30-year fixed-rate mortgages granted during the week of July 6 to July
10, 2009, had the following rates.
4.80
5.47
6.36
5.95
5.46
6.18
5.16
5.78
4.67
5.61
5.65
4.97
5.75
5.62
6.17
5.07
5.43
4.65
5.05
5.10
5.06
5.83
5.39
6.09
5.01
5.85
5.10
5.93
5.21
4.80
5.41
4.67
5.58
5.50
5.36
5.54
5.54
5.93
5.84
5.43
5.28
4.74
4.89
5.83
5.86
6.19
4.97
4.73
5.48
5.98
496 Chapter 10 Estimation and Hypothesis Testing: Two Populations
Another 45 randomly selected 20-year fixed-rate mortgages granted during the week of July 6 to July 10,
2009, had the following rates.
5.64
5.53
4.91
5.63
5.09
5.25
5.18
5.04
5.32
5.63
5.08
5.03
5.27
5.9
5.34
5.41
4.58
4.81
4.76
4.94
5.54
5.62
5.4
4.89
5.9
5.57
4.68
5.03
5.5
5.72
4.74
5.28
4.98
6.52
5.56
4.99
5.57
5.36
5.85
4.91
5.86
4.79
5.66
5.56
4.61
Assume that the population standard deviations of the interest rates on all 30-year fixed-rate mortgages
and all 20-year fixed rate mortgages are the same.
a. Construct a 99% confidence interval for the difference between the mean mortgage rates on all
30-year fixed-rate mortgages and all 20-year fixed-rate mortgages for the said period.
b. Test at the 5% significance level whether the average rate on all 30-year fixed-rate mortgages granted
during the week of July 6 to July 10, 2009, was higher than the average rate on all 20-year fixed-rate mort-
gages granted during the same time period.
TA10.2 A company recently opened two supermarkets in two different areas. The management wants to
know if the mean sales per day for these two supermarkets are different. A sample of 10 days for Super-
market A produced the following data on daily sales (in thousand dollars).
47.56 57.66 51.23 58.29 43.71
49.33 52.35 50.13 47.45 53.86
A sample of 12 days for Supermarket B produced the following data on daily sales (in thousand dollars).
56.34 63.55 61.64 63.75 54.78 58.19
55.40 59.44 62.33 67.82 56.65 67.90
Assume that the daily sales of the two supermarkets are both normally distributed with equal but unknown
standard deviations.
a. Construct a 99% confidence interval for the difference between the mean daily sales for these two
supermarkets.
b. Test at the 1% significance level whether the mean daily sales for these two supermarkets are
different.
TA10.3 Refer to Technology Assignment TA 10.1. Now do that assignment without assuming that the
population standard deviations are the same.
TA10.4 Refer to Technology Assignment TA10.2. Now do that assignment assuming the daily sales of
the two supermarkets are both normally distributed with unequal and unknown standard deviations.
TA10.5 As was mentioned in Exercise 9.38, The Bath Heritage Days, which take place in Bath, Maine,
switched to a Whoopie Pie eating contest in 2009. Suppose the contest involves eating nine Whoopie Pies,
each weighing 1/3 pound. The following data represent the times (in seconds) taken by each contestant
to eat the first Whoopie Pie and the ninth (last) Whoopie Pie. Thirteen contestants actually finished all
nine Whoopie Pies.
Contestant
1
2
3
4
5
6
7
8
9
10
11
12
13
First pie
49
59
66
49
63
70
77
59
64
69
60
58
71
Last pie
49
74
92
93
91
73
103
59
85
94
84
87
111
a. Make a 95% confidence interval for the mean of the population paired differences, where a paired
difference is equal to the time needed to eat the ninth pie minus the time needed to eat the first pie.
b. Using the 10% significance level, can you conclude that it takes at least 15 more seconds, on aver-
age, to eat the ninth pie than to eat the first pie?
Assume that the population of paired differences is (approximately) normally distributed.
TA10.6 A company is considering installing new machines to assemble its products. The company is
considering two types of machines, but it will buy only one type. The company selected eight assembly
workers and asked them to use these two types of machines to assemble products. The following table
gives the time taken (in minutes) to assemble one unit of the product on each type of machine for each of
these eight workers.
Technology Assignments 497
Machine I
23
26
19
24
27
22
20
18
Machine II
21
24
23
25
24
28
24
23
a. Construct a 98% confidence interval for the mean /x d of the population paired differences, where a
paired difference is equal to the time taken to assemble a unit of the product on Machine I minus the time
taken to assemble a unit of the product on Machine II by the same worker.
b. Test at the 5% significance level whether the mean time taken to assemble a unit of the product is
different for the two types of machines.
Assume that the population of paired differences is (approximately) normally distributed.
TA10.7 A company has two restaurants in two different areas of New York City. The company wants
to estimate the percentages of patrons who think that the food and service at each of these restaurants
are excellent. A sample of 200 patrons taken from the restaurant in Area A showed that 118 of them
think that the food and service are excellent at this restaurant. Another sample of 250 patrons selected
from the restaurant in Area B showed that 160 of them think that the food and service are excellent at
this restaurant.
a. Construct a 97% confidence interval for the difference between the two population proportions.
b. Testing at the 2.5% significance level, can you conclude that the proportion of patrons at the restau-
rant in Area A who think that the food and service are excellent is lower than the corresponding propor-
tion at the restaurant in Area B?
TA10.8 The management of a supermarket wanted to investigate whether the percentages of all men
and all women who prefer to buy national brand products over the store brand products are different.
A sample of 600 men shoppers at the company's supermarkets showed that 246 of them prefer to buy
national brand products over the store brand products. Another sample of 700 women shoppers at the
company's supermarkets showed that 266 of them prefer to buy national brand products over the store
brand products.
a. Construct a 99% confidence interval for the difference between the proportions of all men and all
women shoppers at these supermarkets who prefer to buy national brand products over the store brand
products.
b. Testing at the 2% significance level, can you conclude that the proportions of all men and all women
shoppers at these supermarkets who prefer to buy national brand products over the store brand products
are different?
Chapter
Chi-Square Tests
11.1 The Chi-Square
Distribution
1 1 .2 A Goodness-of-Fit Test
Case Study 11-1 What Is
Your Favorite Season?
1 1 .3 Contingency Tables
1 1 .4 A Test of Independence
or Homogeneity
1 1 .5 Inferences About the
Population Variance
Is Winter you favorite season? If yes, you are part of the small percentage of Americans whose
favorite season is winter. But if your response is no, then you have a lot of company. Which sea-
son is the most favorite for people? In a recent survey, Americans mentioned Spring and Fall as
their favorite seasons with 38% and 28% of them picking these seasons, respectively. It may not
surprise you that Winter was preferred by a mere 6% (See Case Study 11-1).
The tests of hypothesis about the mean, the difference between two means, the proportion, and the
difference between two proportions were discussed in Chapters 9 and 10. The tests about propor-
tions dealt with countable or categorical data. In the case of a proportion and the difference between
two proportions in Chapters 9 and 10, the tests concerned experiments with only two categories. Re-
call from Chapter 5 that such experiments are called binomial experiments.
This chapter describes three types of tests:
1. Tests of hypothesis for experiments with more than two categories, called good ness-of -fit tests
2. Tests of hypothesis about contingency tables, called independence and homogeneity tests
3. Tests of hypothesis about the variance and standard deviation of a single population
All of these tests are performed by using the chi-square distribution, which is sometimes written
as x 2 distribution and is read as "chi-square distribution." The symbol x is the Greek letter chi, pro-
nounced "ki." The values of a chi-square distribution are denoted by the symbol x 2 ( rea d as "chi-
square"), just as the values of the standard normal distribution and the f distribution are denoted by
z and f, respectively. Section 11.1 describes the chi-square distribution.
498
11.1 The Chi-Square Distribution 499
11.1 The Chi-Square Distribution
Like the t distribution, the chi-square distribution has only one parameter, called the degrees of
freedom {df). The shape of a specific chi-square distribution depends on the number of degrees
of freedom. 1 (The degrees of freedom for a chi-square distribution are calculated by using dif-
ferent formulas for different tests. This will be explained when we discuss those tests.) The ran-
dom variable % 2 assumes nonnegative values only. Hence, a chi-square distribution curve starts
at the origin (zero point) and lies entirely to the right of the vertical axis. Figure 11.1 shows
three chi-square distribution curves. They are for 2, 7, and 12 degrees of freedom, respectively.
Figure 11.1 Three chi-square
distribution curves.
012345678 9 10 11 12 13 14 15 16 1718 19 2021
As we can see from Figure 11.1, the shape of a chi-square distribution curve is skewed for
very small degrees of freedom, and it changes drastically as the degrees of freedom increase.
Eventually, for large degrees of freedom, the chi-square distribution curve looks like a normal
distribution curve. The peak (or mode) of a chi-square distribution curve with 1 or 2 degrees of
freedom occurs at zero and for a curve with 3 or more degrees of freedom at df — 2. For in-
stance, the peak of the chi-square distribution curve with df = 2 in Figure 11.1 occurs at zero.
The peak for the curve with df = 7 occurs at 7 — 2 = 5. Finally, the peak for the curve with
df = 12 occurs at 12 — 2 = 10. Like all other continuous distribution curves, the total area un-
der a chi-square distribution curve is 1 .0.
Definition
The Chi-Square Distribution The chi-square distribution has only one parameter, called the
degrees of freedom. The shape of a chi-square distribution curve is skewed to the right for small
df and becomes symmetric for large df. The entire chi-square distribution curve lies to the right
of the vertical axis. The chi-square distribution assumes nonnegative values only, and these are
denoted by the symbol x~ (read as "chi-square").
If we know the degrees of freedom and the area in the right tail of a chi-square distribu-
tion curve, we can find the value of x 2 from Table VI of Appendix C. Examples 1 1-1 and 1 1-2
show how to read that table.
■ EXAMPLE 11-1
Find the value of x 1 f° r 7 degrees of freedom and an area of .10 in the right tail of the chi-
square distribution curve.
Solution To find the required value of x 2 , we locate 7 in the column for df and .100 in
the top row in Table VI of Appendix C. The required x 2 value is given by the entry at the
Reading the chi-square
distribution table: area in the
right tail known.
'The mean of a chi-square distribution is equal to its df, and the standard deviation is equal to V2 df.
500 Chapter 1 1 Chi-Square Tests
intersection of the row for 7 and the column for .100. This value is 12.017. The relevant
portion of Table VI is presented as Table 11.1 here.
Table 11.1 x 1 f( > r 4f=7 and .10 Area in the Right Tail
Area in the Right Tail Under the Chi-Square Distribution Curve
df
.995 ••• | .100 1
.005
1
2
0.000 2.706
0.010 • • • 4.605
7.879
10.597
0.989 •■• 1 12.017 1 <
20.278
100
67.328 • ■ • 118.498
140.169
Required value of x 2
As shown in Figure 11.2, the x value for df = 1 and an area of .10 in the right tail of the
chi-square distribution curve is 12.017.
Figure 11.2
12.017 X 2 j
■ EXAMPLE 11-2
Find the value of x 1 for 12 degrees of freedom and an area of .05 in the left tail of the chi-
square distribution curve.
Solution We can read Table VI of Appendix C only when an area in the right tail of the
chi-square distribution curve is known. When the given area is in the left tail, as in this exam-
ple, the first step is to find the area in the right tail of the chi-square distribution curve as fol-
lows.
Area in the right tail = 1 — Area in the left tail
Therefore, for our example,
Area in the right tail = 1 - .05 = .95
Next, we locate 12 in the column for df and .950 in the top row in Table VI of Appendix C.
The required value of given by the entry at the intersection of the row for 12 and the col-
umn for .950, is 5.226. The relevant portion of Table VI is presented as Table 11.2 here.
Reading the chi-square
distribution table: area in the
left tail known.
11.1 The Chi-Square Distribution 501
Table 11.2 x 2 f< > r d f = 12 a nd - 95 Ar ea in the Right Tail
df
Area in the Right Tail Under the Chi-Square Distribution Curve
.950
.005
100
0.004
0.103
5.226
77.929
7.879
10.597
28.300
140.169
Required value of \~
As shown in Figure 11.3, the x 2 value for df = 12 and .05 area in the left tail is 5.226.
Figure 11.3
5.226
EXERCISES
CONCEPTS AND PROCEDURES
11.1 Describe the chi-square distribution. What is the parameter (parameters) of such a distribution?
11.2 Find the value of x 2 for 12 degrees of freedom and an area of .025 in the right tail of the chi-square
distribution curve.
11.3 Find the value of x 2 for 28 degrees of freedom and an area of .05 in the right tail of the chi-square
distribution curve.
11.4 Determine the value of x 2 f° r 14 degrees of freedom and an area of .10 in the left tail of the chi-
square distribution curve.
11.5 Determine the value of x 2 for 23 degrees of freedom and an area of .990 in the left tail of the chi-
square distribution curve.
11.6 Find the value of x 2 for 4 degrees of freedom and
a. .005 area in the right tail of the chi-square distribution curve
b. .05 area in the left tail of the chi-square distribution curve
11.7 Determine the value of x 2 for 13 degrees of freedom and
a. .025 area in the left tail of the chi-square distribution curve
b. .995 area in the right tail of the chi-square distribution curve
502
Chapter 1 1 Chi-Square Tests
1 1 .2 A Goodness-of-Fit Test
This section explains how to make tests of hypothesis about experiments with more than
two possible outcomes (or categories). Such experiments, called multinomial experiments,
possess four characteristics. Note that a binomial experiment is a special case of a multino-
mial experiment.
Definition
A Multinomial Experiment An experiment with the following characteristics is called a multi-
nomial experiment.
1. It consists of n identical trials (repetitions).
2. Each trial results in one of k possible outcomes (or categories), where k > 2.
3. The trials are independent.
4. The probabilities of the various outcomes remain constant for each trial.
An experiment of many rolls of a die is an example of a multinomial experiment. It con-
sists of many identical rolls (trials); each roll (trial) results in one of the six possible outcomes;
each roll is independent of the other rolls; and the probabilities of the six outcomes remain con-
stant for each roll.
As a second example of a multinomial experiment, suppose we select a random sample of
people and ask them whether or not the quality of American cars is better than that of Japanese
cars. The response of a person can be yes, no, or does not know. Each person included in the
sample can be considered as one trial (repetition) of the experiment. There will be as many tri-
als for this experiment as the number of persons selected. Each person can belong to any of the
three categories — yes, no, or does not know. The response of each selected person is independ-
ent of the responses of other persons. Given that the population is large, the probabilities of a
person belonging to the three categories remain the same for each trial. Consequently, this is an
example of a multinomial experiment.
The frequencies obtained from the actual performance of an experiment are called the ob-
served frequencies. In a goodness-of-fit test, we test the null hypothesis that the observed fre-
quencies for an experiment follow a certain pattern or theoretical distribution. The test is called
a goodness-of-fit test because the hypothesis tested is how good the observed frequencies fit a
given pattern.
For our first example involving the experiment of many rolls of a die, we may test the null
hypothesis that the given die is fair. The die will be fair if the observed frequency for each out-
come is close to one-sixth of the total number of rolls.
For our second example involving opinions of people on the quality of American cars,
suppose such a survey was conducted in 2009, and in that survey 41% of the people said yes,
48% said no, and 11% said do not know. We want to test if these percentages still hold true.
Suppose we take a random sample of 1000 adults and observe that 536 of them think that the
quality of American cars is better than that of Japanese cars, 362 say it is worse, and 102 have
no opinion. The frequencies 536, 362, and 102 are the observed frequencies. These frequen-
cies are obtained by actually performing the survey. Now, assuming that the 2009 percentages
are still true (which will be our null hypothesis), in a sample of 1000 adults we will expect
410 to say yes, 480 to say no, and 110 to say do not know. These frequencies are obtained by
multiplying the sample size (1000) by the 2009 proportions. These frequencies are called the
expected frequencies. Then, we will make a decision to reject or not to reject the null hypoth-
esis based on how large the difference between the observed frequencies and the expected fre-
quencies is. To perform this test, we will use the chi-square distribution. Note that in this case
we are testing the null hypothesis that all three percentages (or proportions) are unchanged.
1 1 .2 A Goodness-of-Fit Test 503
However, if we want to make a test for only one of the three proportions, we use the proce-
dure learned in Section 9.4 of Chapter 9. For example, if we are testing the hypothesis that the
current percentage of people who think the quality of American cars is better than that of the
Japanese cars is different from 41%, then we will test the null hypothesis H : p = .41 against
the alternative hypothesis H x : p ¥= .41. This test will be conducted using the procedure dis-
cussed in Section 9.4 of Chapter 9.
As mentioned earlier, the frequencies obtained from the performance of an experiment are
called the observed frequencies. They are denoted by O. To make a goodness-of-fit test, we cal-
culate the expected frequencies for all categories of the experiment. The expected frequency for
a category, denoted by E, is given by the product of n and p, where n is the total number of tri-
als and p is the probability for that category.
Definition
Observed and Expected Frequencies The frequencies obtained from the performance of an
experiment are called the observed frequencies and are denoted by O. The expected frequencies,
denoted by E, are the frequencies that we expect to obtain if the null hypothesis is true. The
expected frequency for a category is obtained as
E = np
where n is the sample size and p is the probability that an element belongs to that category if the
null hypothesis is true.
Degrees of Freedom for a Goodness-of-Fit Test In a goodness-of-fit test, the degrees of freedom
are
df=k-l
where Ic denotes the number of possible outcomes (or categories) for the experiment.
The procedure to make a goodness-of-fit test involves the same five steps that we used in
the preceding chapters. The chi-square goodness-of-fit test is always a right-tailed test.
Test Statistic for a Goodness-of-Fit Test The test statistic for a goodness-of-fit test is x 2 , and its
value is calculated as
2 = y (g ~ E f
x 2j E
where O = observed frequency for a category
E = expected frequency for a category = np
Remember that a chi-square goodness-of-fit test is always a right-tailed test.
Whether or not the null hypothesis is rejected depends on how much the observed and ex-
pected frequencies differ from each other. To find how large the difference between the observed
frequencies and the expected frequencies is, we do not look at just 2(<9 — E), because some of
the O — E values will be positive and others will be negative. The net result of the sum of
these differences will always be zero. Therefore, we square each of the O — E values to obtain
(O — E) 2 , and then we weight them according to the reciprocals of their expected frequencies.
The sum of the resulting numbers gives the computed value of the test statistic x 2 -
504 Chapter 1 1 Chi-Square Tests
To make a goodness-of-fit test, the sample size should be large enough so that the expected
frequency for each category is at least 5. If there is a category with an expected frequency of
less than 5, either increase the sample size or combine two or more categories to make each ex-
pected frequency at least 5.
Examples 1 1-3 and 1 1-4 describe the procedure for performing goodness-of-fit tests us-
ing the chi-square distribution.
Conducting a goodness-
of-fit test: equal proportions
for all categories.
■ EXAMPLE 11-3
A bank has an ATM installed inside the bank, and it is available to its customers only from
7 am to 6 pm Monday through Friday. The manager of the bank wanted to investigate if the
number of transactions made on this ATM are the same for each of the 5 days (Monday through
Friday) of the week. She randomly selected one week and counted the number of transactions
made on this ATM on each of the 5 days during this week. The information she obtained is
given in the following table, where the number of users represents the number of transactions
on this ATM on these days. For convenience, we will refer to these transactions as "people"
or "users."
Day
Monday
Tuesday
Wednesday
Thursday
Friday
Number of users
253
197
204
279
267
At the 1% level of significance, can we reject the null hypothesis that the number of people
who use this ATM each of the 5 days of the week is the same? Assume that this week is typ-
ical of all weeks in regard to the use of this ATM.
Solution To conduct this test of hypothesis, we proceed as follows.
Step 1. State the null and alternative hypotheses.
Because there are 5 categories (days) as listed in the table, the number of ATM users will
be the same for each of these 5 days if 20% of all users use the ATM each day. The null and
alternative hypotheses are as follows.
H : The number of people using the ATM is the same for all 5 days of the week.
H x : The number of people using the ATM is not the same for all 5 days of the week.
If the number of people using this ATM is the same for all 5 days of the week, then
.20 of the users will use this ATM on any of the 5 days of the week. Let p x , p 2 , Pz, Pa,
and p s be the proportions of people who use this ATM on Monday, Tuesday, Wednesday,
Thursday, and Friday, respectively. Then, the null and alternative hypotheses can also be
written as
H - Pi = Pi = Pi = P\ = Ps = -20
Hji At least two of the five proportions are not equal to .20
Step 2. Select the distribution to use.
Because there are 5 categories (i.e., 5 days on which the ATM is used), this is a multino-
mial experiment. Consequently, we use the chi-square distribution to make this test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is given to be .01, and the goodness-of-fit test is always right-tailed.
Therefore, the area in the right tail of the chi-square distribution curve is
Area in the right tail = a = .01
The degrees of freedom are calculated as follows:
k = number of categories = 5
of/ = £-1=5-1=4
1 1 .2 A Goodness-of-Fit Test 505
From the chi-square distribution table (Table VI of Appendix C), the critical value of \ 2 f° r
df = 4 and .01 area in the right tail of the chi-square distribution curve is 13.277, as shown
in Figure 1 1 .4.
Step 4. Calculate the value of the test statistic.
Table 11.3
Calculating the Value of the Test Statistic
Observed
Expected
Category
Frequency
Frequency
(O - E) 2
(Day)
P
E = np
(O -E)
(O - Ef
E
Monday
253
.20
1200(.20) = 240
13
169
.704
Tuesday
197
.20
1200(.20) = 240
-43
1849
7.704
Wednesday
204
.20
1200020) = 240
-36
1296
5.400
Thursday
279
.20
1200020) = 240
39
1521
6.338
Friday
267
.20
1200020) = 240
27
729
3.038
n = 1200
Sum = 23.184
All the required calculations to find the value of the test statistic \ 2 are shown in Table 11.3.
The calculations made in Table 11.3 are explained next.
1. The first two columns of Table 11.3 list the 5 categories (days) and the observed fre-
quencies for the sample of 1200 persons who used the ATM during each of the 5 days
of the selected week. The third column contains the probabilities for the 5 categories
assuming that the null hypothesis is true.
2. The fourth column contains the expected frequencies. These frequencies are obtained
by multiplying the sample size (« = 1200) by the probabilities listed in the third col-
umn. If the null hypothesis is true (i.e., the ATM users are equally distributed over all
5 days), then we will expect 240 out of 1200 persons to use the ATM each day. Con-
sequently, each category in the fourth column has the same expected frequency.
3. The fifth column lists the differences between the observed and expected frequencies,
that is, O — E. These values are squared and recorded in the sixth column.
4. Finally, we divide the squared differences (that appear in the sixth column) by the cor-
responding expected frequencies (listed in the fourth column) and write the resulting
numbers in the seventh column.
506 Chapter 1 1 Chi-Square Tests
5. The sum of the seventh column gives the value of the test statistic x 2 - Thus,
X
2 w = 23.184
Conducting a goodness-
of-fit test: testing if results of a
survey fit a given distribution.
Step 5. Make a decision.
The value of the test statistic x 2 = 23.184 is larger than the critical value of x 2 = 13.277,
and it falls in the rejection region. Hence, we reject the null hypothesis and state that the num-
ber of persons who use this ATM is not the same for the 5 days of the week. In other words,
we conclude that a higher number of users of this ATM use this machine on one or more of
these days.
If you make this chi-square test using any of the statistical software packages, you will ob-
tain a /?-value for the test. In this case you can compare the p-value obtained in the computer
output with the level of significance and make a decision. As you know from Chapter 9, you
will reject the null hypothesis if a (significance level) is greater than the /7-value and not reject
it otherwise.
■ EXAMPLE 11-4
In a July 23, 2009, Harris Interactive Poll, 1015 advertisers were asked about their opinions
of Twitter. The percentage distribution of their responses is shown in the following table.
Opinion
Percentage
A: Twitter is something that is just at its infancy, and its use will grow exponentially 45
over the next few years
B: Twitter is something that mostly young people and the media will use, but it will 21
not move more into the mainstream
C: Twitter is already over, and it is time to find the next best thing 17
D: I do not know enough about Twitter to have an opinion 17
Source: http://www.harrisinteractive.com/harris poll/pubs/Harris Poll 2009 07 23.pdf.
Assume that these percentages hold true for the 2009 population of advertisers. Recently 800
randomly selected advertisers were asked the same question. The following table lists the num-
ber of advertisers in this sample who gave each response.
Opinion
Frequency
A: Twitter is something that is just at its infancy, and its use will grow
exponentially over the next few years
B: Twitter is something that mostly young people and the media will use,
but it will not move more into the mainstream
C: Twitter is already over, and it is time to find the next best thing
D: I do not know enough about Twitter to have an opinion
374
183
127
116
Test at the 2.5% level of significance whether the current distribution of opinions is different
from that for 2009.
Solution We perform the following five steps for this test of hypothesis.
Step 1. State the null and alternative hypotheses.
The null and alternative hypotheses are
H : The current percentage distribution of opinions is the same as for 2009.
H x : The current percentage distribution of opinions is different from that for 2009.
1 1 .2 A Goodness-of-Fit Test 507
Step 2. Select the distribution to use.
Because this experiment has four categories as listed in the table, it is a multinomial ex-
periment. Consequently we use the chi-square distribution to make this test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is given to be .025, and because the goodness-of-fit test is always
right-tailed, the area in the right tail of the chi-square distribution curve is
Area in the right tail = a = .025
The degrees of freedom are calculated as follows:
k = number of categories = 4
#=£-1=4-1=3
From the chi-square distribution table (Table VI of Appendix C), the critical value of
for df = 3 and .025 area in the right tail of the chi-square distribution curve is 9.348, as shown
in Figure 11.5.
Do not reject H — >■
-< — Reject H
r a= .025
and df = 3
Figure IT. 5
9.348 x 2
t
Critical value of x 2
Step 4. Calculate tlie value of the test statistic.
All the required calculations to find the value of the test statistic )f are shown in Table
11.4. Note that the four categories are denoted by A, B, C, and D, respectively, and the per-
centages for 2009 have been converted into probabilities and recorded in the third column of
Table 1 1 .4. The value of the test statistic )f is given by the sum of the last column. Thus,
^ (O - Ef
X 2 = 2^^ = 6-420
Table 11.4 Calculating the Value of the Test Statistic
Observed Expected
Category
Frequency
Frequency
(O - Ef
(Opinion)
O
P
E = np
(O-E)
(O - E) 2
E
No
374
.45
800(.45) = 360
14
196
.544
Yes
183
.21
800(.21) = 168
15
225
1.339
Yes
127
.17
800(.17) = 136
-9
81
.596
Somewhat
116
.17
800(.17) = 136
-20
400
2.941
n = 800
Sum = 5.420
Step 5. Make a decision.
The value of the test statistic x 2 = 5.420 is smaller than the critical value of y 2 = 9.348,
and it falls in the nonrejection region. Hence, we fail to reject the null hypothesis, and state
that the current percentage distribution of opinions is the same as for 2009.
USA TODAY Snapshots®
An outbreak of spring fever
38% Percentage of Americans who
say this is their favorite season:
28% 27%
6%
Spring Fall Summer Winter
Source: SrrategyOne survey for Back to
Nature foods of 1.002 adults Fel> 5-9
By Anne R. Carey and Julie Snider, USA TODAY
The accompanying chart shows the percentage distribution of the opinions of a sample of adults about
their favorite seasons. As the chart shows, 38% of adults in the sample said that Spring is their favorite sea-
son, 28% indicated Fall, 27% mentioned Summer, and only 6% mentioned Winter as their favorite season.
If we add these percentages, they add up to 99%. They do not add up to 100% because of rounding. Sup-
pose that, to make the sum equal to 100%, we change the percentage of adults who liked Spring from
38% to 39%. Assume that these percentages are true for the population of adults at the time of the sur-
vey (February 2009). Suppose that we want to test the hypothesis that these percentages with regard to
the favorite seasons of adults are still true. Then the two hypotheses are as follows:
H : The current percentage distribution of favorite seasons is the same as for 2009.
H,: The current percentage distribution of favorite seasons is not the same as for 2009.
To test this hypothesis, suppose that we take a sample of 1000 adults and ask them about their fa-
vorite season. Suppose that 350 of them say Spring, 320 say Fall, 290 say Summer, and 40 mention Win-
ter. Using the given information, we calculate the value of the test statistic as shown in the following
table.
Observed
Expected
(O - E) 2
Frequency
Frequency
(O - E) 2
Category
O
P
E = np
(O -£)
E
Spring
350
.39
1000(39) = 390
-40
1600
4.103
Fall
320
.28
1000(.28) = 280
40
1600
5.714
Summer
290
.27
1000(.27) = 270
20
400
1.481
Winter
40
.06
1000(.06) = 60
-20
400
6.667
n = 1000
Sum = 17.965
Suppose that we use a 5% significance level to perform this test. Then for df = 4 - 1 =3 and .05
area in the right tail, the observed value of x 2 from Table VI in Appendix C is 7.815. Thus, we will reject the
null hypothesis if the observed value of x 2 is 7.815 or larger. From the above table, the observed value of
X 2 = 17.965 is larger than the critical value of x 2 = 7.815. Consequently, we reject the null hypothesis. Thus
we can conclude that the current percentage distribution of adults with regard to their favorite season is
different from the one for 2009.
508
WHAT IS
YOUR
FAVORITE
SEASON?
Source: Chart reproduced with
permission from USA TODAY, March 20,
2009. Copyright © 2009, USA TODAY.
1 1 .2 A Goodness-of-Fit Test 509
If you make this chi-square test using any of the statistical software packages, you will ob-
tain a /rvalue for the test. In this case you can compare the /7-value obtained in the computer
output with the level of significance and make a decision. As you know from Chapter 9, you
will reject the null hypothesis if a (significance level) is greater than the /7-value and not re-
ject it otherwise. H
EXERCISES
CONCEPTS AND PROCEDURES
11.8 Describe the four characteristics of a multinomial experiment.
11.9 What is a goodness-of-fit test and when is it applied? Explain.
11.10 Explain the difference between the observed and expected frequencies for a goodness-of-fit test.
11.11 How is the expected frequency of a category calculated for a goodness-of-fit test? What are the de-
grees of freedom for such a test?
11.12 To make a goodness-of-fit test, what should be the minimum expected frequency for each category?
What are the alternatives if this condition is not satisfied?
11.13 The following table lists the frequency distribution for 60 rolls of a die.
Outcome
1-spot
2- spot
3-spot
4-spot
5-spot
6-spot
Frequency
7
12
8
15
11
7
Test at the 5% significance level whether the null hypothesis that the given die is fair is true.
■ APPLICATIONS
11.14 In June 2009, the Gallup-Healthways Weil-Being Index (well-beingindex.com) reported that 49%
of Americans exercise for 30 minutes or more on 2 or fewer days per week, 24% exercise for 30 minutes
or more on 3 or 4 days per week, and 27% exercise for 30 minutes or more on 5 or more days per week.
Assume that the Gallup-Healthways results were true for the June 2009 population of Americans. Sup-
pose a recent random sample of 458 Americans produced the following results.
2 or fewer
3 or 4 days
5 or more
Exercise frequency
days a week
a week
days a week
Number of people
197
121
140
Test at the 10% significance level whether the current distribution of exercise frequency differs from that
of June 2009.
11.15 In May 2009, a GFK Roper poll asked Americans about their preferences about attending a
wedding based on the food being served, given a set of choices. Of the respondents, 15.46% said that
they would prefer to go to a wedding that serves champagne and caviar, 58.76% would prefer a wed-
ding that serves wine and chicken breasts, 20.62% would like to go to a wedding that serves beer and
pigs in a blanket (miniature hot dogs wrapped in crescent rolls), and 5.15% were not able to choose.
Assume that these percentages were true for the population of Americans at the time of the poll. Sup-
pose that a recent poll asked the same question of 600 randomly selected Americans, which produced
the following results.
Champagne
Wine and chicken
Beer and pigs
Response
and caviar
breasts
in a blanket
Not sure
Frequency
104
294
184
18
Test at the 5% significance level whether the distribution of responses for the recent sample differs from
that of May 2009.
11.16 Clasp your hands together. Which thumb is on top? Believe it or not, the thumb that you place on
top is determined by genetics. If either of your parents has the gene that tells you to place your left thumb
on top and passes it on to you, you will place your left thumb on top. The left-thumb gene is called the
510 Chapter 11 Chi-Square Tests
dominant gene, which means that if either parent passes it on to you, you will have that trait. If you place
your right thumb on top, you received the recessive gene from both parents. If both parents have both the
left and right thumb genes (the case denoted Lr), Mendelian genetics gives the probabilities listed in the
following table about the children's genes.
Lr (left-thumbed, but also received
Child's genes
LL (left-thumbed)
a right-thumbed gene)
rr (right-thumbed)
Probability
.25
.50
.25
Source: http://humangenetics.suitel01.com/article.cfm/dominant_human_genetic_traits.
Suppose that a random sample of 65 children whose both parents had Lr genes were tested for the genes.
The following table lists the results of this experiment.
Child's genes
LL
Lr
rr
Frequency
14
31
20
Test at the 5% significance level whether the genes received by the sample of children are significantly
different from what Mendelian genetics predicts.
11.17 A drug company is interested in investigating whether the color of their packaging has any im-
pact on sales. To test this, they used five different colors (blue, green, orange, red, and yellow) for the
boxes of an over-the-counter pain reliever, instead of their traditional white box. The following table shows
the number of boxes of each color sold during the first month.
Box color
Blue
Green
Orange
Red
White
Number of boxes sold
310
292
280
216
296
Using the 1% significance level, test the null hypothesis that the number of boxes sold of each of these
five colors is the same.
11.18 Over the last 3 years, Art's Supermarket has observed the following distribution of modes of pay-
ment in the express lines: cash (C) 41%, check (CK) 24%, credit or debit card (D) 26%, and other (N)
9%. In an effort to make express checkout more efficient, Art's has just begun offering a 1% discount for
cash payment in the express checkout line. The following table lists the frequency distribution of the modes
of payment for a sample of 500 express-line customers after the discount went into effect.
Mode of payment
C
CK
D
N
Number of customers
240
104
111
45
Test at the 1% significance level whether the distribution of modes of payment in the express checkout
line changed after the discount went into effect.
11.19 Home Mail Corporation sells products by mail. The company's management wants to find out if
the number of orders received at the company's office on each of the 5 days of the week is the same. The
company took a sample of 400 orders received during a 4-week period. The following table lists the fre-
quency distribution for these orders by the day of the week.
Day of the week
Mon
Tue
Wed
Thu
Fri
Number of orders received
92
71
65
83
89
Test at the 5% significance level whether the null hypothesis that the orders are evenly distributed over all
days of the week is true.
11.20 Of all students enrolled at a large undergraduate university, 19% are seniors, 23% are juniors, 27%
are sophomores, and 31% are freshmen. A sample of 200 students taken from this university by the student
senate to conduct a survey includes 50 seniors, 46 juniors, 55 sophomores, and 49 freshmen. Using the 10%
significance level, test the null hypothesis that this sample is a random sample. (Hint: This sample will be a
random sample if it includes approximately 19% seniors, 23% juniors, 27% sophomores, and 31% freshmen.)
11.21 Chance Corporation produces beauty products. Two years ago the quality control department at the
company conducted a survey of users of one of the company's products. The survey revealed that 53% of
the users said the product was excellent, 31% said it was satisfactory, 7% said it was unsatisfactory, and
1 1 .4 A Test of Independence or Homogeneity 51 1
9% had no opinion. Assume that these percentages were true for the population of all users of this product
at that time. After this survey was conducted, the company redesigned this product. A recent survey of 800
users of the redesigned product conducted by the quality control department at the company showed that
495 of the users think the product is excellent, 255 think it is satisfactory, 35 think it is unsatisfactory, and
15 have no opinion. Is the percentage distribution of the opinions of users of the redesigned product differ-
ent from the percentage distribution of users of this product before it was redesigned? Use a = .025.
11.22 Henderson Corporation makes metal sheets, among other products. When the process that is used
to make metal sheets works properly, 92% of the metal sheets contain no defects, 5% have one defect
each, and 3% have two or more defects each. The quality control inspectors at the company take samples
of metal sheets quite often and check them for defects. If the distribution of defects for a sample is sig-
nificantly different from the above-mentioned percentage distribution, the process is stopped and adjusted.
A recent sample of 300 sheets produced the frequency distribution of defects listed in the following table.
Number of defects
None
One
Two or more
Number of metal sheets
262
24
14
Does the evidence from this sample suggest that the process needs an adjustment? Use a = .01.
11.3 Contingency Tables
Often we may have information on more than one variable for each element. Such information
can be summarized and presented using a two-way classification table, which is also called a
contingency table or cross-tabulation. Suppose a university has a total of 20,758 students en-
rolled. By classifying these students based on gender and whether these students are full-time
or part-time, we can prepare Table 11.5, which provides an example of a contingency table.
Table 11.5 has two rows (one for males and the second for females) and two columns (one for
full-time and the second for part-time students). Hence, it is also called a 2 X 2 (read as "two
by two") contingency table.
Table 11.5 Total Enrollment at a University
Full-time
Part-time
Male
6768
2615 <-
Female
7658
3717
Students who are
male and enrolled
part-time
A contingency table can be of any size. For example, it can be 2 X 3, 3 X 2, 3 X 3, or
4X2. Note that in these notations, the first digit refers to the number of rows in the table,
and the second digit refers to the number of columns. For example, a 3 X 2 table will con-
tain three rows and two columns. In general, an R X C table contains R rows and C columns.
Each of the four boxes that contain numbers in Table 11.5 is called a cell. The number of
cells in a contingency table is obtained by multiplying the number of rows by the number of
columns. Thus, Table 1 1.5 contains 2X2 = 4 cells. The subjects that belong to a cell of a con-
tingency table possess two characteristics. For example, 2615 students listed in the second cell
of the first row in Table 11.5 are male and part-time. The numbers written inside the cells are
usually called the joint frequencies. For example, 2615 students belong to the joint category of
male and part-time. Hence, it is referred to as the joint frequency of this category.
11.4 A Test of Independence or Homogeneity
This section is concerned with tests of independence and homogeneity, which are performed
using contingency tables. Except for a few modifications, the procedure used to make such tests
is almost the same as the one applied in Section 11.2 for a goodness-of-fit test.
512 Chapter 11 Chi-Square Tests
1 1 .4.1 A Test of Independence
In a test of independence for a contingency table, we test the null hypothesis that the two at-
tributes (characteristics) of the elements of a given population are not related (that is, they are
independent) against the alternative hypothesis that the two characteristics are related (that is,
they are dependent). For example, we may want to test if the affiliation of people with the De-
mocratic and Republican parties is independent of their income levels. We perform such a test
by using the chi-square distribution. As another example, we may want to test if there is an as-
sociation between being a man or a woman and having a preference for watching sports or soap
operas on television.
Definition
Degrees of Freedom for a Test of Independence A test of independence involves a test of the
null hypothesis that two attributes of a population are not related. The degrees of freedom for a
test of independence are
df=(R- 1)(C- 1)
where R and C are the number of rows and the number of columns, respectively, in the given
contingency table.
The value of the test statistic x 2 m a test of independence is obtained using the same for-
mula as in the goodness-of-fit test described in Section 1 1 .2.
Test Statistic for a Test of Independence The value of the test statistic x 2 fo r a test of independ-
ence is calculated as
where O and E are the observed and expected frequencies, respectively, for a cell.
The null hypothesis in a test of independence is always that the two attributes are not re-
lated. The alternative hypothesis is that the two attributes are related.
The frequencies obtained from the performance of an experiment for a contingency table
are called the observed frequencies. The procedure to calculate the expected frequencies for
a contingency table for a test of independence is different from the one for a goodness-of-fit
test. Example 1 1-5 describes this procedure.
Calculating expected
frequencies for a test of
independence.
■ EXAMPLE 11-5
Violence and lack of discipline have become major problems in schools in the United States.
A random sample of 300 adults was selected, and these adults were asked if they favor giv-
ing more freedom to schoolteachers to punish students for violence and lack of discipline. The
two-way classification of the responses of these adults is presented in the following table.
In Favor
Against
No Opinion
CP)
(A)
(AO
Men (M)
93
70
12
Women (W)
87
32
6
1 1 .4 A Test of Independence or Homogeneity 513
Calculate the expected frequencies for this table, assuming that the two attributes, gender and
opinions on the issue, are independent.
Solution The preceding table is reproduced as Table 1 1.6 here. Note that Table 1 1.6 includes
the row and column totals.
Table 11.6 Observed Frequencies
In Favor
Against
No Opinion
Row
(F)
(A)
(AO
Totals
Men (M)
93
70
12
175
Women (W)
87
32
6
125
Column Totals
180
102
18
300
The numbers 93, 70, 12, 87, 32, and 6 listed inside the six cells of Table 11.6 are called
the observed frequencies of the respective cells.
As mentioned earlier, the null hypothesis in a test of independence is that the two attrib-
utes (or classifications) are independent. In an independence test of hypothesis, first we as-
sume that the null hypothesis is true and that the two attributes are independent. Assuming
that the null hypothesis is true and that gender and opinions are not related in this example,
we calculate the expected frequency for the cell corresponding to Men and In Favor as shown
next. From Table 11.6,
P(a person is a Man) = P(M) = 175/300
P{& person is In Favor) = P(F) = 180/300
Because we are assuming that M and F are independent (by assuming that the null hypothesis is
true), from the formula learned in Chapter 4, the joint probability of these two events is
P(M and F) = P(M) X P(F) = (175/300) X (180/300)
Then, assuming that M and F are independent, the number of persons expected to be Men and
In Favor in a sample of 300 is
E for Men and In Favor = 300 X P(M and F)
175 180 175 X 180
= 300 X X =
300 300 300
(Row total)(Column total)
Sample size
Thus, the rule for obtaining the expected frequency for a cell is to divide the product of the
corresponding row and column totals by the sample size.
Expected Frequencies for
a Test of Independence The expected frequency E for a cell is
calculated as
(Row total) (Column total)
E =
Sample size
Using this rule, we calculate the expected frequencies of the six cells of Table 11.6 as follows:
E for Men and In Favor cell = (175)(180)/300 = 105.00
E for Men and Against cell = (175)(102)/300 = 59.50
E for Men and No Opinion cell = (175)(18)/300 = 10.50
514 Chapter 11 Chi-Square Tests
E for Women and In Favor cell = (125)(180)/300 = 75.00
E for Women and Against cell = (125)(102)/300 = 42.50
£for Women and No Opinion cell = (125)(18)/300 = 7.50
The expected frequencies are usually written in parentheses below the observed frequen-
cies within the corresponding cells, as shown in Table 11.7.
Table 11.7 Observed and Expected Frequencies
In Favor
Against
No Opinion
Row
(F)
(A)
(AO
Totals
Men (M)
93
70
12
175
(105.00)
(59.50)
(10.50)
Women (W)
87
32
6
125
(75.00)
(42.50)
(7.50)
Column Totals
180
102
18
300
Like a goodness-of-fit test, a test of independence is always right-tailed. To apply a chi-
square test of independence, the sample size should be large enough so that the expected
frequency for each cell is at least 5. If the expected frequency for a cell is not at least 5, we
either increase the sample size or combine some categories. Examples 11-6 and 11-7 de-
scribe the procedure to make tests of independence using the chi-square distribution.
■ EXAMPLE 11-6
Reconsider the two-way classification table given in Example 1 1-5. In that example, a ran-
dom sample of 300 adults was selected, and they were asked if they favor giving more free-
dom to schoolteachers to punish students for violence and lack of discipline. Based on the
results of the survey, a two-way classification table was prepared and presented in Exam-
ple 11-5. Does the sample provide sufficient evidence to conclude that the two attributes,
gender and opinions of adults, are dependent? Use a 1% significance level.
Solution The test involves the following five steps.
Step 1. State the null and alternative hypotheses.
As mentioned earlier, the null hypothesis must be that the two attributes are independent.
Consequently, the alternative hypothesis is that these attributes are dependent.
H : Gender and opinions of adults are independent.
Hi. Gender and opinions of adults are dependent.
Step 2. Select the distribution to use.
We use the chi-square distribution to make a test of independence for a contingency table.
Step 3. Determine the rejection and nonrejection regions.
The significance level is 1%. Because a test of independence is always right-tailed, the area
of the rejection region is .01, and it falls in the right tail of the chi-square distribution curve.
The contingency table contains two rows (Men and Women) and three columns (In Favor,
Against, and No Opinion). Note that we do not count the row and column of totals. The de-
grees of freedom are
df= (R - 1)(C- 1) = (2 - 1)(3 - 1) = 2
Making a test of
independence: 2x3
table.
1 1 .4 A Test of Independence or Homogeneity
From Table VI of Appendix C, the critical value of x 2 f° r df = 2 and a = .01 is 9.210. This
value is shown in Figure 11.6.
Step 4. Calculate the value of the test statistic.
Table 11.7, with the observed and expected frequencies constructed in Example 11-5, is
reproduced as Table 11.8.
Table T1.8 Observed and Expected Frequencies
In Favor
Against
No Opinion
Row
(F)
(A)
m
Totals
Men (M)
93
70
12
175
(105.00)
(59.50)
(10.50)
Women (W)
87
32
6
125
(75.00)
(42.50)
(7.50)
Column Totals
180
102
18
300
To compute the value of the test statistic x 2 , we take the difference between each pair of ob-
served and expected frequencies listed in Table 1 1 .8, square those differences, and then divide
each of the squared differences by the respective expected frequency. The sum of the resulting
numbers gives the value of the test statistic x 2 - All these calculations are made as follows:
x 2
^ E
(93 - 105.00) 2 + (70 - 59.50) 2 (12 - 10.50) 2
105.00 59.50 10.50
^ (87 - 75.00) 2 (32 - 42.50) 2 (6 - 7.50) 2
75.00 42.50 7.50
= 1.371 + 1.853 + .214 + 1.920 + 2.594 + .300 = 8.252
Step 5. Make a decision.
The value of the test statistic x 2 = 8.252 is less than the critical value of x 2 = 9.210, and
it falls in the nonrejection region. Hence, we fail to reject the null hypothesis and state that
there is not enough evidence from the sample to conclude that the two characteristics, gender
and opinions of adults, are dependent for this issue. I
■ EXAMPLE 11-7
A researcher wanted to study the relationship between gender and owning cell phones. She
took a sample of 2000 adults and obtained the information given in the following table.
Making a test of
independence: 2x2
table.
516 Chapter 11 Chi-Square Tests
Own Cell Phones
Do Not Own Cell Phones
Men
640
450
Women
440
470
At the 5% level of significance, can you conclude that gender and owning a cell phone are re-
lated for all adults?
Solution We perform the following five steps to make this test of hypothesis.
Step 1. State the null and alternative hypotheses.
The null and alternative hypotheses are, respectively,
H : Gender and owning a cell phone are not related.
H\. Gender and owning a cell phone are related.
Step 2. Select the distribution to use.
Because we are performing a test of independence, we use the chi-square distribution to
make the test.
Step 3. Determine the rejection and nonrejection regions.
With a significance level of 5%, the area of the rejection region is .05, and it falls into the
right tail of the chi-square distribution curve. The contingency table contains two rows (men
and women) and two columns (own cell phones and do not own cell phones). The degrees of
freedom are
df= (R - 1)(C- 1) = (2 - 1)(2 - 1) = 1
From Table VI of Appendix C, the critical value of x 1 f° r 4f = 1 and a
value is shown in Figure 11.7.
.05 is 3.841. This
Step 4. Calculate the value of the test statistic.
The expected frequencies for the various cells are calculated as follows, and as listed within
parentheses in Table 11.9.
Table 11.9 Observed and Expected Frequencies
Own Cell Phones
Do Not Own Cell Phones
Row
(Y)
(AO
Totals
Men (M)
640
450
1090
(588.60)
(501.40)
Women (W)
440
470
910
(491.40)
(418.60)
Column Totals 1080 920 2000
1 1 .4 A Test of Independence or Homogeneity 517
E for men and own cell phones cell = (1090)(1080)/2000 = 588.60
E for men and do not own cell phones cell = (1090)(920)/2000 = 501.40
E for women and own cell phones cell = (910)(1080)/2000 = 491.40
E for women and do not own cell phones cell = (910)(920)/2000 = 418.60
The value of the test statistic \ 2 is calculated as follows:
(640 - 588.60) 2 (450 - 501.40) 2 (440 - 491.40) 2 (470 - 418.60) 2
588.60 501.40 491.40 418.60
= 4.489 + 5.269 + 5.376 + 6.311 = 21.445
Step 5. Make a decision.
The value of the test statistic \ 2 = 21.445 is larger than the critical value of x 2 = 3.841,
and it falls into the rejection region. Hence, we reject the null hypothesis and state that there
is strong evidence from the sample to conclude that the two characteristics, gender and own-
ing cell phones, are related for all adults. I
1 1 .4.2 A Test of Homogeneity
In a test of homogeneity, we test if two (or more) populations are homogeneous (similar) with re-
gard to the distribution of a certain characteristic. For example, we might be interested in testing
the null hypothesis that the proportions of households that belong to different income groups are
the same in California and Wisconsin, or we may want to test whether or not the preferences of
people in Florida, Arizona, and Vermont are similar with regard to Coke, Pepsi, and 7-Up.
Definition
A Test of Homogeneity A test of homogeneity involves testing the null hypothesis that the pro-
portions of elements with certain characteristics in two or more different populations are the
same against the alternative hypothesis that these proportions are not the same.
Let us consider the example of testing the null hypothesis that the proportions of house-
holds in California and Wisconsin who belong to various income groups are the same. (Note
that in a test of homogeneity, the null hypothesis is always that the proportions of elements with
certain characteristics are the same in two or more populations. The alternative hypothesis is
that these proportions are not the same.) Suppose we define three income strata: high-income
group (with an income of more than $150,000), medium-income group (with an income of
$60,000 to $150,000), and low-income group (with an income of less than $60,000). Further-
more, assume that we take one sample of 250 households from California and another sample
of 150 households from Wisconsin, collect the information on the incomes of these households,
and prepare the contingency table as in Table 11.10.
Table 11.10 Observed Frequencies „
Row
California
Wisconsin
Totals
High income
70
34
104
Medium income
80
40
120
Low income
100
76
176
Column Totals 250 150 400
518 Chapter 11 Chi-Square Tests
Note that in this example the column totals are fixed. That is, we decided in advance to
take samples of 250 households from California and 150 from Wisconsin. However, the row to-
tals (of 104, 120, and 176) are determined randomly by the outcomes of the two samples. If we
compare this example to the one about violence and lack of discipline in schools in the previ-
ous section, we will notice that neither the column nor the row totals were fixed in that exam-
ple. Instead, the researcher took just one sample of 300 adults, collected the information on gen-
der and opinions, and prepared the contingency table. Thus, in that example, the row and column
totals were all determined randomly. Thus, when both the row and column totals are determined
randomly, we perform a test of independence. However, when either the column totals or the
row totals are fixed, we perform a test of homogeneity. In the case of income groups in Cali-
fornia and Wisconsin, we will perform a test of homogeneity to test for the similarity of income
groups in the two states.
The procedure to conduct a test of homogeneity is similar to the procedure used to make
a test of independence discussed earlier. Like a test of independence, a test of homogeneity is
right-tailed. Example 1 1-8 illustrates the procedure to make a homogeneity test.
■ EXAMPLE 11-8
Consider the data on income distributions for households in California and Wisconsin given
in Table 11.10. Using the 2.5% significance level, test the null hypothesis that the distri-
bution of households with regard to income levels is similar (homogeneous) for the two
states.
Solution We perform the following five steps to make this test of hypothesis.
Step 1. State the null and alternative hypotheses.
The two hypotheses are, respectively, 2
H : The proportions of households that belong to different income groups are the
same in both states.
Hji The proportions of households that belong to different income groups are not
the same in both states.
Step 2. Select the distribution to use.
We use the chi-square distribution to make a homogeneity test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is 2.5%. Because the homogeneity test is right-tailed, the area of the
rejection region is .025, and it lies in the right tail of the chi-square distribution curve. The
contingency table for income groups in California and Wisconsin contains three rows and two
columns. Hence, the degrees of freedom are
df=(R- 1)(C- 1) = (3- 1)(2- 1) = 2
From Table VI of Appendix C, the value of x 1 f° r 4f = 2 and .025 area in the right tail of the
chi-square distribution curve is 7.378. This value is shown in Figure 11.8.
2 Let p HC , p MC , and p LC be the proportions of households in California who belong to high-, middle-, and low-income
groups, respectively. Let p HW , Pmw> an d Plw be the corresponding proportions for Wisconsin. Then we can also write
the null hypothesis as
Ha'-Pnc = Phw. Pmc = Pmw> an d Plc = Plw
and the alternative hypothesis as
Hj: At least two of the equalities mentioned in H are not true.
Performing a test of
homogeneity.
1 1 .4 A Test of Independence or Homogeneity 519
Step 4. Calculate the value of the test statistic.
To compute the value of the test statistic )f, we need to calculate the expected frequencies
first. Table 11.11 lists the observed and the expected frequencies. The numbers in parenthe-
ses in this table are the expected frequencies, which are calculated using the formula
(Row total) (Column total)
E =
Total of both samples
Thus, for instance,
(104)(250)
E for High income and California cell = = 65
6 J 400
Table 11.11 Observed and Expected Frequencies
Row
California
Wisconsin
Totals
High income
70
(65)
34
(39)
104
Medium income
80
(75)
40
(45)
120
Low income
100
(110)
76
(66)
176
Column Totals 250 150 400
The remaining expected frequencies are calculated in the same way. Note that the expected
frequencies in a test of homogeneity are calculated in the same way as in a test of independ-
ence. The value of the test statistic x 2 is computed as follows:
2 ^ (° ~ E f
x =2:—^-
(70 - 65) 2 (34 - 39) 2 (80 - 75) 2 (40 - 45) 2
65 + 39 + 75 + 45
(100 - 110) 2 (76 - 66) 2
+ " 110 " + 66
= .385 + .641 + .333 + .556 + .909 + 1.515 = 4.339
Step 5. Make a decision.
The value of the test statistic \ 2 = 4.339 is less than the critical value of x 2 = 7.378, and
it falls in the nonrejection region. Hence, we fail to reject the null hypothesis and state that
the distribution of households with regard to income appears to be similar (homogeneous) in
California and Wisconsin.
520 Chapter 1 1 Chi-Square Tests
EXERCISES
CONCEPTS AND PROCEDURES
11.23 Describe in your own words a test of independence and a test of homogeneity. Give one example
of each.
11.24 Explain how the expected frequencies for cells of a contingency table are calculated in a test of in-
dependence or homogeneity. How do you find the degrees of freedom for such tests?
11.25 To make a test of independence or homogeneity, what should be the minimum expected frequency
for each cell? What are the alternatives if this condition is not satisfied?
11.26 Consider the following contingency table, which is based on a sample survey.
Column 1
Column 2
Column 3
Row 1
137
64
105
Row 2
98
71
65
Row 3
115
81
115
a. Write the null and alternative hypotheses for a test of independence for this table.
b. Calculate the expected frequencies for all cells, assuming that the null hypothesis is true.
c. For a = .01, find the critical value of x 2 - Show the rejection and nonrejection regions on the
chi-square distribution curve.
d. Find the value of the test statistic \ .
e. Using a = .01, would you reject the null hypothesis?
11.27 Consider the following contingency table, which records the results obtained for four samples of
fixed sizes selected from four populations.
Sample Selected From
Population 1
Population 2
Population 3
Population 4
Row 1
24
81
60
121
Row 2
46
64
91
72
Row 3
20
37
105
93
a. Write the null and alternative hypotheses for a test of homogeneity for this table.
b. Calculate the expected frequencies for all cells assuming that the null hypothesis is true.
c. For a = .025, find the critical value of x 2 - Show the rejection and nonrejection regions on the
chi-square distribution curve.
d. Find the value of the test statistic x 2 -
e. Using a = .025, would you reject the null hypothesis?
■ APPLICATIONS
11.28 During the recent economic recession, many families faced hard times financially. Some studies ob-
served that more people stopped buying name brand products and started buying less expensive store brand
products instead. Data produced by a recent sample of 700 adults on whether they usually buy store brand
or name brand products are recorded in the following table.
More Often Buy
Name Brand
Store Brand
Men
150
165
Women
160
225
Using the 1% significance level, can you reject the null hypothesis that the two attributes, gender and buy-
ing name or store brand products, are independent?
1 1 .4 A Test of Independence or Homogeneity 521
11.29 One hundred auto drivers who were stopped by police for some violation were also checked to see
if they were wearing seat belts. The following table records the results of this survey.
Wearing Seat Belt
Not Wearing Seat Belt
Men
40
15
Women
38
7
Test at the 2.5% significance level whether being a man or a woman and wearing or not wearing a seat
belt are related.
11.30 Many students graduate from college deeply in debt from student loans, credit card debts, and so
on. A sociologist took a random sample of 401 single persons, classified them by gender, and asked,
"Would you consider marrying someone who was $25,000 or more in debt?" The results of this survey
are shown in the following table.
Yes
No
Uncertain
Women
125
59
21
Men
101
79
16
Test at the 1% significance level whether gender and response are related.
11.31 During the Bush and Obama administrations, there has been a great deal of discussion about partisan-
ship. Did partisanship have an impact on approval ratings in public polls? The following tables display the ap-
proval ratings of both presidents during April of their first year in office. The first table shows the approval
numbers for each president from voters of his own party (Democrats for Obama, Republicans for Bush). The
second table shows the approval numbers for each president from voters of the opposition party (Republicans
for Obama, Democrats for Bush). The numbers are comparable to the percentages reported by gallup.com.
President's Own Party President's Opposition Party
Obama
(April 2009)
Bush
(April 2001)
Obama
(April 2009)
Bush
(April 2001)
Approve
1091
1046
Approve
335
430
Not sure/
Not sure/
disapprove
109
154
disapprove
865
770
a. Test at the 1% significance level whether the approval ratings by their own party voters are related.
b. Test at the 1 % significance level whether the approval ratings by the opposition party voters are
related.
11.32 The game show Deal or No Deal involves a series of opportunities for the contestant to either ac-
cept an amount of money from the show's banker or to decline it and open a specific number of brief-
cases in the hope of exposing and, thereby eliminating, low amounts of money from the game, which
would lead the banker to increase the amount of the next offer. Suppose that 700 people aged 21 years
and older were selected at random. Each of them watched an episode of the show until exactly four brief-
cases were left unopened. The money amounts in these four briefcases were $750, $5000, $50,000, and
$400,000, respectively. The banker's offer to the contestant was $81,600 if the contestant would stop the
game and accept the offer. If the contestant were to decline the offer, he or she would choose one brief-
case out of these four to open, and then there would be a new offer. All 700 persons were asked whether
they would accept the offer (Deal) for $81,600 or turn it down (No Deal), as well as their ages. The re-
sponses of these 700 persons are listed in the following table.
Age Group (years)
21-29
30-39
40-49
50-59
60 and over
Deal
78
82
89
92
63
No Deal
56
70
60
63
47
Test at the 5% significance level whether the decision to accept or not to accept the offer (Deal or No Deal)
and age group are dependent.
522 Chapter 1 1 Chi-Square Tests
11.33 A forestry official is comparing the causes of forest fires in two regions, A and B. The following
table shows the causes of fire for 76 randomly selected recent fires in these two regions.
Arson
Accident
Lightning
Unknown
Region A
6
9
6
10
Region B
7
14
15
9
Test at the 5% significance level whether causes of fire and regions of fires are related.
11.34 National Electronics Company buys parts from two subsidiaries. The quality control department
at this company wanted to check if the distribution of good and defective parts is the same for the sup-
plies of parts received from both subsidiaries. The quality control inspector selected a sample of 300
parts received from Subsidiary A and a sample of 400 parts received from Subsidiary B. These parts
were checked for being good or defective. The following table records the results of this investigation.
Subsidiary A
Subsidiary B
Good
284
381
Defective
16
19
Using the 5% significance level, test the null hypothesis that the distributions of good and defective parts
are the same for both subsidiaries.
11.35 Two drugs were administered to two groups of randomly assigned 60 and 40 patients, respec-
tively, to cure the same disease. The following table gives information about the number of patients who
were cured and not cured by each of the two drugs.
Cured
Not Cured
Drug I
44
16
Drug II
18
22
Test at the 1% significance level whether or not the two drugs are similar in curing and not curing the patients.
11.36 Four hundred people were selected from each of the four geographic regions (Midwest, Northeast,
South, West) of the United States, and they were asked which form of camping they prefer. The choices
were pop-up camper/trailer, family style (tenting with sanitary facilities), rustic (tenting, no sanitary facil-
ities), or none. The results of the survey are shown in the following table.
Midwest
Northeast
South
West
Camper/trailer
132
129
129
135
Family style
180
175
168
146
Rustic
46
50
59
68
None
42
46
44
51
Based on the evidence from these samples, can you conclude that the distributions of favorite forms of
camping are different for at least two of the regions? Use a = .01.
11.37 A FOX News/Opinion Dynamics poll asked a question about gun control of random samples of
900 people each during May 2009 and March 2000. The question asked was, "Which of the following do
you think is more likely to decrease gun violence: better enforcement of existing gun laws or more laws
and restrictions on obtaining guns?" The numbers in the following table are approximately the same as
reported in the poll, which was reported to the nearest percent.
Better
More Laws
Enforcement
and Restrictions
Both
Unsure
May 2009
425
308
93
74
March 2000
372
330
122
76
Source: http://www.pollingreport.com/guns.htm.
Test at the 5% significance level whether the distributions of responses from May 2009 and March 2000
are significantly different.
1 1 .5 Inferences About the Population Variance 523
11.38 The following table gives the distributions of grades for three professors for a few randomly se-
lected classes that each of them taught during the last 2 years.
Professor
Miller
Smith
Moore
A
18
36
20
B
25
44
15
Grade
C
85
73
82
D & F
17
12
8
Using the 2.5% significance level, test the null hypothesis that the grade distributions are homogeneous
for these three professors.
11.39 Two random samples, one of 95 blue-collar workers and a second of 50 white-collar workers, were
taken from a large company. These workers were asked about their views on a certain company issue. The
following table gives the results of the survey.
Opinion
Favor
Oppose
Uncertain
Blue-collar workers
44
39
12
White-collar workers
21
26
3
Using the 2.5% significance level, test the null hypothesis that the distributions of opinions are homoge-
neous for the two groups of workers.
11.5 Inferences About the Population Variance
Earlier chapters explained how to make inferences (confidence intervals and hypothesis tests)
about the population mean and population proportion. However, we may often need to control
the variance (or standard deviation). Consequently, there may be a need to estimate and to test
a hypothesis about the population variance a 2 . Section 11.5.1 describes how to make a confi-
dence interval for the population variance (or standard deviation). Section 11.5.2 explains how
to test a hypothesis about the population variance.
As an example, suppose a machine is set up to fill packages of cookies so that the net
weight of cookies per package is 32 ounces. Note that the machine will not put exactly 32 ounces
of cookies into each package. Some of the packages will contain less and some will contain
more than 32 ounces. However, if the variance (and, hence, the standard deviation) is too large,
some of the packages will contain quite a bit less than 32 ounces of cookies, and some others
will contain quite a bit more than 32 ounces. The manufacturer will not want a large variation
in the amounts of cookies put into different packages. To keep this variation within some spec-
ified acceptable limit, the machine will be adjusted from time to time. Before the manager de-
cides to adjust the machine at any time, he or she must estimate the variance or test a hypoth-
esis or do both to find out if the variance exceeds the maximum acceptable value.
Like every sample statistic, the sample variance is a random variable, and it possesses a
sampling distribution. If all the possible samples of a given size are taken from a population
and their variances are calculated, the probability distribution of these variances is called the
sampling distribution of the sample variance.
Sampling Distribution of (n — ])s 2 /<r 2 If the population from which the sample is selected is
(approximately) normally distributed, then
(« - l)s 2
2
a'
has a chi-square distribution with n — 1 degrees of freedom.
524 Chapter 1 1 Chi-Square Tests
Thus, the chi-square distribution is used to construct a confidence interval and test a hy-
pothesis about the population variance cr 2 .
11.5.1 Estimation of the Population Variance
The value of the sample variance s 2 is a point estimate of the population variance cr 2 . The
(1 — a) 100% confidence interval for a 2 is given by the following formula.
Confidence Interval for the Population Variance a 1 Assuming that the population from which
the sample is selected is (approximately) normally distributed, we obtain the (1 — a) 100% con-
fidence interval for the population variance cr 2 as
(n - l)s 2 (n - l)s 2
2 to 2
Xa/2 Xl-a/2
where Xa/2 an d X\-a/2 are obtained from the chi-square distribution table for a/2 and
1 — a/2 areas in the right tail of the chi-square distribution curve, respectively, and for
n — 1 degrees of freedom.
The confidence interval for the population standard deviation can be obtained by simply taking
the positive square roots of the two limits of the confidence interval for the population variance.
The procedure for making a confidence interval for cr 2 involves the following three steps.
1. Take a sample of size n and compute s 2 using the formula learned in Chapter 3. However,
if n and s 2 are given, then perform only steps 2 and 3.
2. Calculate a/2 and 1 — a/2. Find two values of \ 2 from the chi-square distribution table
(Table VI of Appendix C): one for a/2 area in the right tail of the chi-square distribution
curve and df = n — 1, and the second for 1 — a/2 area in the right tail and df = n — 1.
3. Substitute all the values in the formula for the confidence interval for o 2 and simplify.
Example 1 1-9 illustrates the estimation of the population variance and population standard
deviation.
■ EXAMPLE 11-9
One type of cookie manufactured by Haddad Food Company is Cocoa Cookies. The machine
that fills packages of these cookies is set up in such a way that the average net weight of these
packages is 32 ounces with a variance of .015 square ounce. From time to time the quality
control inspector at the company selects a sample of a few such packages, calculates the vari-
ance of the net weights of these packages, and constructs a 95% confidence interval for the
population variance. If either both or one of the two limits of this confidence interval is not
in the interval .008 to .030, the machine is stopped and adjusted. A recently taken random
sample of 25 packages from the production line gave a sample variance of .029 square ounce.
Based on this sample information, do you think the machine needs an adjustment? Assume
that the net weights of cookies in all packages are normally distributed.
Solution The following three steps are performed to estimate the population variance and
to make a decision.
Step 1. From the given information, n = 25 and s 2 = .029
Step 2. The confidence level is 1 — a = .95. Hence, a = 1 — .95 = .05. Therefore,
a/2 = .05/2 = .025
1 - a/2 = 1 - .025 = .975
df = n - 1 = 25 - 1 = 24
Constructing confidence
intervals for cr 2 and <j.
1 1 .5 Inferences About the Population Variance 525
From Table VI of Appendix C,
X 2 for 24 df and .025 area in the right tail = 39.364
X 2 for 24 df and .975 area in the right tail = 12.401
These values are shown in Figure 11.9.
dfm 24
Figure 11.9
,-2= .025
39.364
L
Value of Xa
Step 3. The 95% confidence interval for a is
in — l)s 2 (n — l)s
2 t0 2
Xa/2 Xl-a/2
(25 - 1)(.029) (25 - 1)(.029)
39.364 t0 12.401
.0177 to .0561
Thus, with 95% confidence, we can state that the variance for all packages of Cocoa Cookies
lies between .0177 and .0561 square ounce. Note that the lower limit (.0177) of this confi-
dence interval is between .008 and .030, but the upper limit (.0561) is larger than .030 and
falls outside the interval .008 to .030. Because the upper limit is larger than .030, we can state
that the machine needs to be stopped and adjusted.
We can obtain the confidence interval for the population standard deviation <j by taking the
positive square roots of the two limits of the above confidence interval for the population vari-
ance. Thus, a 95% confidence interval for the population standard deviation is
VXH77 to VX)56T or .133 to .237
Hence, the standard deviation of all packages of Cocoa Cookies is between .133 and .237
ounce at a 95% confidence level.
1 1 .5.2 Hypothesis Tests About the Population Variance
A test of hypothesis about the population variance can be one-tailed or two-tailed. To make
a test of hypothesis about a 2 , we perform the same five steps we used earlier in hypothesis-
testing examples. The procedure to test a hypothesis about cr 2 discussed in this section is ap-
plied only when the population from which a sample is selected is (approximately) normally
distributed.
526 Chapter 1 1 Chi-Square Tests
Test Statistic for a Test of Hypothesis About a- 2 The value of the test statistic x 1 is calculated as
where s 2 is the sample variance, a 2 is the hypothesized value of the population variance, and
n — 1 represents the degrees of freedom. The population from which the sample is selected is
assumed to be (approximately) normally distributed.
Examples 1 1-10 and 1 1-1 1 illustrate the procedure for making tests of hypothesis about cr 2
Performing a right-tailed test
of hypothesis about a 1 .
■ EXAMPLE 11-10
One type of cookie manufactured by Haddad Food Company is Cocoa Cookies. The machine
that fills packages of these cookies is set up in such a way that the average net weight of these
packages is 32 ounces with a variance of .015 square ounce. From time to time the quality con-
trol inspector at the company selects a sample of a few such packages, calculates the variance
of the net weights of these packages, and makes a test of hypothesis about the population vari-
ance. She always uses a = .01. The acceptable value of the population variance is .015 square
ounce or less. If the conclusion from the test of hypothesis is that the population variance is not
within the acceptable limit, the machine is stopped and adjusted. A recently taken random sam-
ple of 25 packages from the production line gave a sample variance of .029 square ounce. Based
on this sample information, do you think the machine needs an adjustment? Assume that the net
weights of cookies in all packages are normally distributed.
Solution From the given information,
n = 25, a = .01, and s 2 = .029
The population variance should not exceed .015 square ounce.
Step 1. State the null and alternative hypotheses.
We are to test whether or not the population variance is within the acceptable limit. The
population variance is within the acceptable limit if it is less than or equal to .015; otherwise,
it is not. Thus, the two hypotheses are
H : cr 2 < .015 (The population variance is within the acceptable limit.)
H\. a 2 > .015 (The population variance exceeds the acceptable limit.)
Step 2. Select the distribution to use.
We use the chi-square distribution to test a hypothesis about cr 2 .
Step 3. Determine the rejection and nonrejection regions.
The significance level is 1% and, because of the > sign in H h the test is right-tailed. The
rejection region lies in the right tail of the chi-square distribution curve with its area equal to
.01. The degrees of freedom for a chi-square test about cr 2 are « — 1; that is,
df = n - 1 = 25 - 1 = 24
From Table VI of Appendix C, the critical value of x 2 for 24 degrees of freedom and .01 area
in the right tail is 42.980. This value is shown in Figure 11.10.
Figure 11.10
Do not reject H -
■ Reject H
Critical value of% 2 -
1 1 .5 Inferences About the Population Variance 527
Step 4. Calculate the value of the test statistic.
The value of the test statistic \ 2 f° r the sample variance is calculated as follows:
2 (n - l)s 2 (25 - 1)(.029)
* = ~ = ^ = 46.400
From H
Step 5. Make a decision.
The value of the test statistic x 2 = 46.400 is greater than the critical value of x 1 = 42.980,
and it falls in the rejection region. Consequently, we reject H and conclude that the popu-
lation variance is not within the acceptable limit. The machine should be stopped and
adjusted. I
■ EXAMPLE 11-11
The variance of scores on a standardized mathematics test for all high school seniors was
150 in 2009. A sample of scores for 20 high school seniors who took this test this year
gave a variance of 170. Test at the 5% significance level if the variance of current scores
of all high school seniors on this test is different from 150. Assume that the scores of all
high school seniors on this test are (approximately) normally distributed.
Solution From the given information,
n = 20, a = .05, and s 2 = 170
The population variance was 150 in 2009.
Step 1. State the null and alternative hypotheses.
The null and alternative hypotheses are, respectively,
Conducting a two-tailed
test of hypothesis about a 1 .
150 (The population variance is not different from 150.)
H{. o 2 ¥= 150 (The population variance is different from 150.)
Step 2. Select the distribution to use.
We use the chi-square distribution to test a hypothesis about a 2 .
Step 3. Determine the rejection and nonrejection regions.
The significance level is 5%. The ¥= sign in H x indicates that the test is two-tailed. The re-
jection region lies in both tails of the chi-square distribution curve with its total area equal to
.05. Consequently, the area in each tail of the distribution curve is .025. The values of a/2
and 1 — a/2 are, respectively,
a
2
.05
.025 and 1
a
2
1
.025 = .975
The degrees of freedom are
df-
20
19
From Table VI of Appendix C, the critical values of x 2 for 19 degrees of freedom and for a/2
and 1 — a/2 areas in the right tail are
X 2 for 19 df and .025 area in the right tail = 32.852
X 2 for 19 fif/and .975 area in the right tail = 8.907
These two values are shown in Figure 11.11.
528
Chapter 1 1 Chi-Square Tests
Step 4. Calculate the value of the test statistic.
The value of the test statistic \ 2 f° r the sample variance is calculated as follows:
, (n - l)s 2 (20 - 1)(170)
X 2 = ~ r- = ~ = 21.533
a 150
From H
Step 5. Make a decision.
The value of the test statistic x 2 = 21.533 is between the two critical values of x 2 , 8.907
and 32.852, and it falls in the nonrejection region. Consequently, we fail to reject H and con-
clude that the population variance of the current scores of high school seniors on this stan-
dardized mathematics test does not appear to be different from 150. I
Note that we can make a test of hypothesis about the population standard deviation a us-
ing the same procedure as that for the population variance a 2 . To make a test of hypothesis
about cr, the only change will be mentioning the values of cr in H and H x . The rest of the pro-
cedure remains the same as in case of cr 2 .
■I
EXERCISES
CONCEPTS AND PROCEDURES
11.40 A sample of certain observations selected from a normally distributed population produced a sam-
ple variance of 46. Construct a 95% confidence interval for cr 2 for each of the following cases and com-
ment on what happens to the confidence interval of cr 2 when the sample size increases.
a. n = 12
b. n = 16
c. n = 25
11.41 A sample of 25 observations selected from a normally distributed population produced a sample
variance of 35. Construct a confidence interval for cr 2 for each of the following confidence levels and com-
ment on what happens to the confidence interval of cr 2 when the confidence level decreases.
a. 1 - a = .99
b. 1 - a = .95
c. 1 - a = .90
11.42 A sample of 22 observations selected from a normally distributed population produced a sample
variance of 18.
a. Write the null and alternative hypotheses to test whether the population variance is different
from 14.
b. Using a = .05, find the critical values of x 2 - Show the rejection and nonrejection regions on a
chi-square distribution curve.
c. Find the value of the test statistic x~-
d. Using the 5% significance level, will you reject the null hypothesis stated in part a?
1 1 .5 Inferences About the Population Variance 529
11.43 A sample of 21 observations selected from a normally distributed population produced a sample
variance of 1.97.
a. Write the null and alternative hypotheses to test whether the population variance is greater
than 1.75.
b. Using a = .025, find the critical value of % 2 . Show the rejection and nonrejection regions on a
chi-square distribution curve.
c. Find the value of the test statistic % 2 .
d. Using the 2.5% significance level, will you reject the null hypothesis stated in part a?
11.44 A sample of 30 observations selected from a normally distributed population produced a sample
variance of 5.8.
a. Write the null and alternative hypotheses to test whether the population variance is different from
6.0.
b. Using a = .05, find the critical value of \ 2 - Show the rejection and nonrejection regions on a
chi-square distribution curve.
c. Find the value of the test statistic \ 2 -
d. Using the 5% significance level, will you reject the null hypothesis stated in part a?
11.45 A sample of 18 observations selected from a normally distributed population produced a sample
variance of 4.6.
a. Write the null and alternative hypotheses to test whether the population variance is different
from 2.2.
b. Using a = .05, find the critical values of \ 2 - Show the rejection and nonrejection regions on a
chi-square distribution curve.
c. Find the value of the test statistic x 2 -
d. Using the 5% significance level, will you reject the null hypothesis stated in part a?
■ APPLICATIONS
11.46 Sandpaper is rated by the coarseness of the grit on the paper. Sandpaper that is more coarse will
remove material faster. Jobs such as the final sanding of bare wood prior to painting or sanding in between
coats of paint require sandpaper that is much finer. A manufacturer of sandpaper rated 220, which is used
for the final preparation of bare wood, wants to make sure that the variance of the diameter of the parti-
cles in their 220 sandpaper does not exceed 2.0 micrometers. Fifty-one randomly selected particles are
measured. The variance of the particle diameters is 2.13 micrometers. Assume that the distribution of par-
ticle diameter is approximately normal.
a. Construct the 95% confidence intervals for the population variance and standard deviation.
b. Test at the 2.5% significance level whether the variance of the particle diameters of all particles
in 220-rated sandpaper is greater than 2.0 micrometers.
11.47 The makers of Flippin' Out Pancake Mix claim that one cup of their mix contains 1 1 grams of
sugar. However, the mix is not uniform, so the amount of sugar varies from cup to cup. One cup of mix
was taken from each of 24 randomly selected boxes. The sample variance of the sugar measurements from
these 24 cups was 1.47 grams. Assume that the distribution of sugar content is approximately normal.
a. Construct the 98% confidence intervals for the population variance and standard deviation.
b. Test at the 1% significance level whether the variance of the sugar content per cup is greater
than 1.0 gram.
11.48 An auto manufacturing company wants to estimate the variance of miles per gallon for its auto
model AST727. A random sample of 22 cars of this model showed that the variance of miles per gallon
for these cars is .62.
a. Construct the 95% confidence intervals for the population variance and standard deviation.
Assume that the miles per gallon for all such cars are (approximately) normally distributed.
b. Test at the 1% significance level whether the sample result indicates that the population vari-
ance is different from .30.
11.49 The manufacturer of a certain brand of lightbulbs claims that the variance of the lives of these bulbs
is 4200 square hours. A consumer agency took a random sample of 25 such bulbs and tested them. The
variance of the lives of these bulbs was found to be 5200 square hours. Assume that the lives of all such
bulbs are (approximately) normally distributed.
a. Make the 99% confidence intervals for the variance and standard deviation of the lives of all
such bulbs.
b. Test at the 5% significance level whether the variance of such bulbs is different from 4200 square
hours.
530 Chapter 1 1 Chi-Square Tests
USES AND MISUSES... DON'T FEED THE ANIMALS
You are a wildlife enthusiast studying African wildlife: gnus, zebras,
and gazelles. You know that a herd of each species will visit one of
three watering places in a region every day, but you do not know the
distribution of choices that the animals make or whether these
choices are dependent. You have observed that the animals some-
times drink together and sometimes do not. A statistician offers to
help and says that he will perform a test for independence of water-
ing place choices based on your observations of the animals' behav-
ior over the past several months. The statistician performs some cal-
culations and says that he has answered your question because his
chi-square test of the independence of watering place choices, at a
5% significance level, told him to reject the null hypothesis. He has
also performed a goodness-of-fit test on the hypothesis that the an-
imals are equally likely to choose any watering place, and he has
rejected that hypothesis as well.
The statistician barely helped you. In the first case, you know a
single piece of information: the choice of a watering place for the
three groups of animals is dependent. Another way of stating the re-
sult is that your data indicate that the choice of watering places for
at least one of the animals is not independent of the others. Perhaps
the zebras get up early, and the gnus and gazelles follow, making the
gnus and gazelles dependent on the choice of the zebras. Or per-
haps the animals choose the watering place of the day independent
of the other animals, but always avoid the watering place at which
the lions are drinking. Regarding the goodness-of-fit test, all you know
is that the hypothesis that the animals equally favor the three water-
ing places was wrong. But you do not know what the expected
distribution should be. In short, the rejection of the null hypothesis
raises more questions than it answers.
Glossary
Chi-square distribution A distribution, with degrees of freedom
as the only parameter, that is skewed to the right for small rf/and
looks like a normal curve for large df.
Expected frequencies The frequencies for different categories of
a multinomial experiment or for different cells of a contingency table
that are expected to occur when a given null hypothesis is true.
Goodness-of-fit test A test of the null hypothesis that the observed
frequencies for an experiment follow a certain pattern or theoretical
distribution.
Multinomial experiment An experiment with n trials for which
(1) the trials are identical, (2) there are more than two possible
outcomes per trial, (3) the trials are independent, and (4) the
probabilities of the various outcomes remain constant for each
trial.
Observed frequencies The frequencies actually obtained from the
performance of an experiment.
Test of homogeneity A test of the null hypothesis that the propor-
tions of elements that belong to different groups in two (or more)
populations are similar.
Test of independence A test of the null hypothesis that two attrib-
utes of a population are not related.
Supplementary Exercises
11.50 According to a report in the Wall Street Journal (http://online.wsj.com/mdc/public/page/
2_3022-autosales.html), the distribution of all auto sales by segment (type of vehicle) in the United States
during June 2009 was as follows.
Segment
Cars
Light-duty trucks
SUVs
Crossovers
June 2009 percentage
42.24
36.72
6.52
14.52
A recent survey of 850 new auto sales produced the following distribution.
Segment
Cars
Light-duty trucks
SUVs
Crossovers
Number of sales
377
299
61
113
Test at the 10% significance level whether the distribution of recent auto sales is significantly different
from the June 2009 distribution.
Supplementary Exercises 531
11.51 One of the products produced by Branco Food Company is Total-Bran Cereal, which competes with
three other brands of similar total-bran cereals. The company's research office wants to investigate if the
percentage of people who consume total-bran cereal is the same for each of these four brands. Let us
denote the four brands of cereal by A, B, C, and D. A sample of 1000 persons who consume total-bran
cereal was taken, and they were asked which brand they most often consume. Of the respondents, 212 said
they usually consume Brand A, 284 consume Brand B, 254 consume Brand C, and 250 consume Brand
D. Does the sample provide enough evidence to reject the null hypothesis that the percentage of people
who consume total-bran cereal is the same for all four brands? Use a = .05.
11.52 The distribution of birth weights (in grams) for all children who shared multiple births (twins,
triplets, etc.) in North Carolina during 2007 was as shown in the table below.
Weight (in grams)
0-500
501-1500
1501-2500
2501-8165
Percentage of children
1.68
12.07
47.39
38.86
The frequency distribution of birth weights of a sample of 200 children who shared multiple births and
were born in North Carolina in 2009 was as shown in the following table.
Weight (in grams)
0-500
501-1500
1501-2500
2501-8165
Frequency distribution
2
22
73
103
Test at the 5% significance level whether the 2009 distribution of birth weights for all children born in
North Carolina who shared multiple births is significantly different from the one for 2007.
11.53 A 2008 online survey by WSOC Channel 9 in Charlotte, North Carolina, asked people to choose
their favorite Christmas song from a list of five choices. The following table shows the percentage of peo-
ple who preferred each of the five songs.
Blue
Carol of
Jingle Bell
White
Winter
Song
Christmas
the Bells
Rock
Christmas
Wonderland
Percentage
5.26
15.45
18.99
39.91
20.39
Source: wsoctv.com/surveyresults/17708782/detail. html?section=holidays&coid=17708782.
Suppose that these percentages are true for the 2008 population of U.S. adults. Suppose that recently 600
randomly selected U.S. adults were asked the same question, and the following table shows the numbers
of adults who preferred each of these five songs.
Blue
Carol of
Jingle Bell
White
Winter
Song
Christmas
the Bells
Rock
Christmas
Wonderland
Number of adults
27
117
102
268
86
Test at the 2.5% significance level whether the distribution of results in the recent survey differs from the
2008 distribution reported by WSOC.
11.54 During a bear market, 140 investors were asked how they were adjusting their portfolios to protect
themselves. Some of these investors were keeping most of their money in stocks, whereas others were
shifting large amounts of money to bonds, real estate, or cash (such as money market accounts). The
results of the survey are shown in the following table.
Favored choice
Stocks
Bonds
Real estate
Cash
Number of investors
46
41
32
21
Using the 2.5% significance level, test the null hypothesis that the percentages of investors favoring the
four choices are all equal.
11.55 A randomly selected sample of 100 persons who suffer from allergies were asked during what sea-
son they suffer the most. The results of the survey are recorded in the following table.
Season
Fall
Winter
Spring
Summer
Persons allergic
18
13
31
38
532 Chapter 1 1 Chi-Square Tests
Using the 1% significance level, test the null hypothesis that the proportions of all allergic persons are
equally distributed over the four seasons.
11.56 All shoplifting cases in the town of Seven Falls are randomly assigned to either Judge Stark or Judge
Rivera. A citizens group wants to know whether either of the two judges is more likely to sentence the of-
fenders to jail time. A sample of 180 recent shoplifting cases produced the following two-way table.
Jail
Other sentence
Judge Stark
27
65
Judge Rivera
31
57
Test at the 5% significance level whether the type of sentence for shoplifting depends on which judge tries
the case.
11.57 A June 2009 CBS News poll asked a random sample of 895 Americans how concerned they were
about their healthcare being impacted and getting worse if the government created a system to provide
healthcare to all Americans. The following table summarizes the results.
Political Affiliation
Democrat Independent
Republican
Very concerned
46 116
88
Somewhat concerned
100 143
61
Not too/not at all concerned
110 188
43
Source: http://www.cbsnews.com/htdocs/pdf/CBSPOLLJune09a_health_care.pdf.
Test at the 1% significance level if political affiliation and concern about one's healthcare quality are
dependent.
11.58 A poll reported in USA TODAY asked American parents if schools should have the primary respon-
sibility for educating teens about drug use. Assuming that the data were based on random samples of 800
fathers and 800 mothers, the percentages reported by USA TODAY would yield the numbers given in the
following table.
School's Responsibility?
Parent
Yes No
Father
238 562
Mother
154 646
Source: http://www.usatoday.com/news/snapshot
.htm?section=N&label=2009-07-14-skuldrug.
Test at the 1% significance level if the responses of fathers and mothers are independent.
11.59 Recent recession and bad economic conditions forced many people to hold more than one job
to make ends meet. A sample of 500 persons who held more than one job produced the following two-
way table.
Single
Married
Other
Male
72
209
39
Female
33
102
45
Test at the 10% significance level whether gender and marital status are related for all people who hold
more than one job.
11.60 ATVs (all-terrain vehicles) have become a source of controversy. Some people feel that their use
should be tightly regulated, while others prefer fewer restrictions. Suppose a survey consisting of a ran-
dom sample of 200 people aged 18 to 27 and another survey of a random sample of 210 people aged 28
to 37 was conducted, and these people were asked whether they favored more restrictions on ATVs, fewer
restrictions, or no change. The results of this survey are summarized in the following table.
Supplementary Exercises 533
More restrictions
Fewer restrictions
No change
18 to 27
40
92
68
Age
28 to 37
55
68
87
Test at the 2.5% significance level whether the distribution of opinions in regard to ATVs are the same for
both age groups.
11.61 A random sample of 100 persons was selected from each of four regions in the United States. These
people were asked whether or not they support a certain farm subsidy program. The results of the survey
are summarized in the following table.
Favor
Oppose
Uncertain
Northeast
56
33
11
Midwest
73
23
4
South
67
28
5
West
59
35
6
Using the 1% significance level, test the null hypothesis that the percentages of people with different opin-
ions are similar for all four regions.
11.62 Construct the 98% confidence intervals for the population variance and standard deviation for the
following data, assuming that the respective populations are (approximately) normally distributed.
a. n = 21, s 2 = 9.2 b. n = 17, s 2 = 1.7
11.63 Construct the 95% confidence intervals for the population variance and standard deviation for the
following data, assuming that the respective populations are (approximately) normally distributed.
a. n = 10, s 2 = 7.2 b. n = 18, s 2 = 14.8
11.64 Refer to Exercise 11.62a. Test at the 5% significance level if the population variance is different
from 6.5.
11.65 Refer to Exercise 11.62b. Test at the 2.5% significance level if the population variance is greater
than 1.1.
11.66 Refer to Exercise 11.63a. Test at the 1% significance level if the population variance is greater
than 4.2.
11.67 Refer to Exercise 11.63b. Test at the 5% significance level if the population variance is different
from 10.4.
11.68 Usually people do not like waiting in line a long time for service. A bank manager does not want the
variance of the waiting times for her customers to be greater than 4.0 square minutes. A random sample of
25 customers taken from this bank gave the variance of the waiting times equal to 8.3 square minutes.
a. Test at the 1% significance level whether the variance of the waiting times for all customers at
this bank is greater than 4.0 square minutes. Assume that the waiting times for all customers
are normally distributed.
b. Construct a 99% confidence interval for the population variance.
11.69 The variance of the SAT scores for all students who took that test this year is 5000. The variance
of the SAT scores for a random sample of 20 students from one school is equal to 3175.
a. Test at the 2.5% significance level whether the variance of the SAT scores for students from
this school is lower than 5000. Assume that the SAT scores for all students at this school are
(approximately) normally distributed.
b. Construct the 98% confidence intervals for the variance and the standard deviation of SAT
scores for all students at this school.
11.70 A company manufactures ball bearings that are supplied to other companies. The machine that is
used to manufacture these ball bearings produces them with a variance of diameters of .025 square
millimeter or less. The quality control officer takes a sample of such ball bearings quite often and checks,
using confidence intervals and tests of hypotheses, whether or not the variance of these bearings is within
.025 square millimeter. If it is not, the machine is stopped and adjusted. A recently taken random sample
of 23 ball bearings gave a variance of the diameters equal to .034 square millimeter.
a. Using the 5% significance level, can you conclude that the machine needs an adjustment?
Assume that the diameters of all ball bearings have a normal distribution.
b. Construct a 95% confidence interval for the population variance.
534 Chapter 1 1 Chi-Square Tests
11.71 A random sample of 25 students taken from a university gave the variance of their GPAs equal to .19.
a. Construct the 99% confidence intervals for the population variance and standard deviation. As-
sume that the GPAs of all students at this university are (approximately) normally distributed.
b. The variance of GPAs of all students at this university was .13 two years ago. Test at the 1%
significance level whether the variance of current GPAs at this university is different from .13.
11.72 A sample of seven passengers boarding a domestic flight produced the following data on weights
(in pounds) of their carry-on bags.
46.3 41.5 39.7 31.0 40.6 35.8 43.2
a. Using the formula from Chapter 3, find the sample variance, s 2 , for these data.
b. Make the 98% confidence intervals for the population variance and standard deviation. Assume
that the population from which this sample is selected is normally distributed.
c Test at the 5% significance level whether the population variance is larger than 20 square
pounds.
11.73 The following are the prices (in dollars) of the same brand of camcorder found at eight stores in
Los Angeles.
568 628 602 642 550 688 615 604
a. Using the formula from Chapter 3, find the sample variance, s 2 , for these data.
b. Make the 95% confidence intervals for the population variance and standard deviation. Assume
that the prices of this camcorder at all stores in Los Angeles follow a normal distribution.
c. Test at the 5% significance level whether the population variance is different from 750 square
dollars.
Advanced Exercises
11.74 A 2009 survey reported in USA TODAY asked U.S. households who cooks in the home on Mother's
Day. The results from the survey are reported in the following table. Assume that these results are true for
the population of all U.S. households in 2009.
Mom with help
Mom's
The guests
Person who cooks
Mom
from the family
spouse
bring food
Percentage of households
38
34
19
9
Source: usatoday.com.
Suppose that recently a random sample of 300 U.S. households were asked the same question and the
number of households with two responses are shown in the following table, and the numbers are missing
for the other two responses.
Mom with help
Mom's
The guests
Person who cooks
Mom
from the family
spouse
bring food
Number of households
114
102
?
?
a. Suppose you were to perform a hypothesis test to compare the sample data to the USA
TODAY percentages. What would the counts for the categories Mom's spouse and The guests
bring food have to be in order for the value of the test statistic to be as small as possible?
Note: There is only one correct pair of values for this question.
b. By how much would the count for Mom 's spouse have to increase from the value in part a in
order to reject the null hypothesis at the 10% significance level?
c. Suppose you were to reduce the count for Mom's spouse in part a by the same amount by
which you increased it in part b. Calculate the value of the test statistic. How does this com-
pare to the value of the test statistic you calculated in part b?
11.75 A chemical manufacturing company wants to locate a hazardous waste disposal site near a city
of 50,000 residents and has offered substantial financial inducements to the city. Two hundred adults
(110 women and 90 men) who are residents of this city are chosen at random. Sixty percent of these adults
oppose the site, 32% are in favor, and 8% are undecided. Of those who oppose the site, 65% are women;
of those in favor, 62.5% are men. Using the 5% level of significance, can you conclude that opinions on
the disposal site are dependent on gender?
11.76 A student who needs to pass an elementary statistics course wonders whether it will make a
difference if she takes the course with instructor A rather than instructor B. Observing the final grades
Supplementary Exercises 535
given by each instructor in a recent elementary statistics course, she finds that Instructor A gave
48 passing grades in a class of 52 students and Instructor B gave 44 passing grades in a class of
54 students. Assume that these classes and grades make simple random samples of all classes and
grades of these instructors.
a. Compute the value of the standard normal test statistic z of Section 10.5.3 for the data and use
it to find the p- value when testing for the difference between the proportions of passing
grades given by these instructors.
b. Construct a 2 X 2 contingency table for these data. Compute the value of the x 2 test statistic
for the test of independence and use it to find the p-value.
c. How do the test statistics in parts a and b compare? How do the p-values for the tests in parts a
and b compare? Do you think this is a coincidence, or do you think this will always happen?
11.77 Each of five boxes contains a large (but unknown) number of red and green marbles. You have been
asked to find if the proportions of red and green marbles are the same for each of the five boxes. You sam-
ple 50 times, with replacement, from each of the five boxes and observe 20, 14, 23, 30, and 18 red marbles,
respectively. Can you conclude that the five boxes have the same proportions of red and green marbles? Use
a .05 level of significance.
11.78 Suppose that you have a two-way table with the following row and column totals.
Variable 1
Total
A
B
C
X
120
Variable 2
Y
205
Z
175
Total
165
140
195
500
The observed values in the cells must be counts, which are nonnegative integers. Calculate the expected
counts for the cells under the assumption that the two variables are independent. Based on your calcula-
tions, explain why it is impossible for the test statistic to have a value of zero.
11.79 You have collected data on a variable, and you want to determine if a normal distribution is a rea-
sonable model for these data. The following table shows how many of the values fall within certain ranges
of z values for these data.
Category Count
z score below —2 48
z score from —2 to less than —1.5 67
z score from —1.5 to less than —1 146
z score from —1 to less than —0.5 248
z score from —0.5 to less than 187
z score from to less than 0.5 125
z score from 0.5 to less than 1 88
z score from 1 to less than 1.5 47
z score from 1.5 to less than 2 25
z score of 2 or above 19
Total 1000
Perform a hypothesis test to determine if a normal distribution is an appropriate model for these data. Use
a significance level of 5%.
11.80 Refer to Problem 11.61. Explain why the hypothesis test in that problem is a test of homogeneity
as opposed to a test of independence. What feature of the data would change if you were to collect data
in order to test for independence?
11.81 You are performing a goodness-of-fit test with four categories, all of which are supposed to be
equally likely. You have a total of 100 observations. The observed frequencies are 21, 26, 31, and 22, re-
spectively, for the four categories.
536 Chapter 1 1 Chi-Square Tests
a. Show that you would fail to reject the null hypothesis for these data for any reasonable signifi-
cance level.
b. The sum of the absolute differences (between the expected and the observed frequencies) for
these data is 14 (i.e., 4+1+6 + 3 = 14). Is it possible to have different observed frequencies
keeping the sum at 14 so that you get a p-value of .10 or less?
Self-Review Test
1. The random variable x 2 assumes only
a. positive b. nonnegative c. nonpositive values
2. The parameter(s) of the chi-square distribution is (are)
a. degrees of freedom b. df and n c. x
3. Which of the following is not a characteristic of a multinomial experiment?
a. It consists of n identical trials.
b. There are k possible outcomes for each trial and k > 2.
c. The trials are random.
d. The trials are independent.
e. The probabilities of outcomes remain constant for each trial.
4. The observed frequencies for a goodness-of-fit test are
a. the frequencies obtained from the performance of an experiment
b. the frequencies given by the product of n and p
c. the frequencies obtained by adding the results of a and b
5. The expected frequencies for a goodness-of-fit test are
a. the frequencies obtained from the performance of an experiment
b. the frequencies given by the product of n and p
c. the frequencies obtained by adding the results of a and b
6. The degrees of freedom for a goodness-of-fit test are
a. n — 1 b. k — 1 c. n + k — 1
7. The chi-square goodness-of-fit test is always
a. two-tailed b. left-tailed c. right-tailed
8. To apply a goodness-of-fit test, the expected frequency of each category must be at least
a. 10 b. 5 c. 8
9. The degrees of freedom for a test of independence are
a. (R - 1)(C - 1) b. n - 2 c. (n - l)(k - 1)
10. According to the Henry J. Kaiser Family Foundation (www.statehealthfacts.org), the percentage dis-
tribution of obesity/overweight people by race/ethnicity in the United States in 2008 was as listed in the
table below.
Race/
ethnicity
African
American
American
Indian/Native
Alaskan
Asian/Pacific
Islander
Hispanic/Latino
(all races)
Other
White
Percentage
15.62
.73
2.86
15.26
.06
65.47
Note: Because of the way the data were presented, the racial categories include all people of those races who are not
Hispanic/Latino.
A recent survey of 20,000 adult Americans who are obese or overweight resulted in the following race/eth-
nicity frequencies.
Race/
ethnicity
African
American
American
Indian/Native
Alaskan
Asian/Pacific
Islander
Hispanic/Latino
(all races)
Other
White
Frequency
3048
137
645
3169
18
12,983
Test at the 5% significance level whether the distribution of race/ethnicity of obese and overweight peo-
ple in the recent survey differs from the 2008 distribution.
Mini-Projects 537
11. The following table gives the two-way classification of 1000 persons who have been married at least
once. They are classified by educational level and marital status.
Educational Level
Less Than
High School
Some
College
High School
Degree
College
Degree
Divorced
173
158
95
53
Never divorced
162
126
110
123
Test at the 1% significance level whether educational level and ever being divorced are dependent.
12. A researcher wanted to investigate if people who belong to different income groups are homogeneous
with regard to playing lotteries. She took a sample of 600 people from the low-income group, another
sample of 500 people from the middle-income group, and a third sample of 400 people from the high-
income group. All these people were asked whether they play the lottery often, sometimes, or never. The
results of the survey are summarized in the following table.
Income Group
Low
Middle
High
Play often
174
163
90
Play sometimes
286
217
120
Never play
140
120
190
Using the 5% significance level, can you reject the null hypothesis that the percentages of people who
play the lottery often, sometimes, and never are the same for each income group?
13. The owner of an ice cream parlor is concerned about consistency in the amount of ice cream his servers
put in each cone. He would like the variance of all such cones to be no more than .25 square ounce. He
decides to weigh each double-dip cone just before it is given to the customer. For a sample of 20 double-
dip cones, the weights were found to have a variance of .48 square ounce. Assume that the weights of all
such cones are (approximately) normally distributed.
a. Construct the 99% confidence intervals for the population variance and the population stan-
dard deviation.
b. Test at the 1% significance level whether the variance of the weights of all such cones ex-
ceeds .25 square ounce.
Mini-Projects
■ MINI-PROJECT 11-1
In recent years drivers have become careless about signaling their turns. To study this problem, go to a
busy intersection and observe at least 75 vehicles that make left turns. Divide these vehicles into three or
four classes. For example, you might use cars, trucks, and others, where "others" include minivans and
sport-utility vehicles, as classes. For each left turn made by a vehicle, record the type of vehicle and whether
or not the driver used the left turn signal before making this turn. It would be better to avoid intersections
that have designated left-turn lanes or green arrows for left turns because drivers in these situations often
assume that their intent to turn left is obvious. Carry out an appropriate test at the 1% level of significance
to determine if signaling behavior and vehicle type are dependent.
■ MINI-PROJECT 11-2
One day during lunch, visit your school cafeteria, observe at least 100 people, and write down what they
are drinking. Categorize the drinks as soft drink (soda, fruit punch, or lemonade), iced tea, milk or juice,
hot drink, and water. Also identify the gender of each person. Perform a hypothesis test to determine if
the type of drink and gender are independent.
538 Chapter 1 1 Chi-Square Tests
■ MINI-PROJECT 11-3
Many studies have been performed to determine the sources that people use to get their news. Survey at
least 50 people at random from your class or dorm and ask the following question:
Which of the following would you classify as being your primary source for news?
a. Network news broadcasts
b. Cable news broadcasts
c. Newspapers
d. Internet-based news sources
e. Radio news broadcasts
Use the data to test the null hypothesis that college students are equally likely to classify the five options
as being their primary source for news. Use a 5% significance level.
■ MINI-PROJECT 11-4
Refer to Case Study 8-2 that discussed the results of a survey in which adults were asked that which of a
set of four noises (car alarm, jackhammer, baby crying, or dog barking) they find to be the most frustrat-
ing to hear. Suppose that the survey also included information on the gender of the respondents as listed
in the following table.
Sound That is Most
Frustrating to Hear
Females
Males
Car alarm
225
197
Jackhammer
154
131
Baby crying
79
143
Dog barking
44
31
a. Perform a hypothesis test to determine whether the sound that is most frustrating to hear and gen-
der are independent. Use a 1% significance level.
b. Suppose that it was noticed that the surveyor incorrectly marked 30 of the survey sheets. Specif-
ically, thirty of the male respondents were mistakenly recorded as saying "baby crying." Further-
more, it was determined that all 30 of the mistakes should have one of the other three responses
but the same response. That is, all these 30 responses should have been "car alarm," or all 30
should have been "jackhammer," or all 30 should have been "dog barking." Determine which of
these three changes would result in a different conclusion than the one you obtained in part a.
DECIDE FOR YOURSELF
Testing for the Fairness of
Gambling Equipment
number of times and observe the frequency for each outcome.
Suppose we roll this die 180 times and obtain the frequencies for
various outcomes as listed in the following table.
Casino gambling has grown rapidly in the United States. Native
American tribes have opened casinos on reservations, many horse
racing tracks have been allowed to add slot machines on site, and
riverboat/lakefront casinos have also been opened in recent years.
States with casino gambling have state agencies that are responsi-
ble for verifying and making sure that the games and equipment
are fair and not fixed. In many states, such an agency is called the
Division of Gaming Enforcement. New Jersey and Nevada have
two of the largest such agencies, given the presence of Atlantic
City and Las Vegas in these states. The chi-square procedures that
you have learned in this chapter can be used to test the validity of
the fairness assumption in regard to the gaming equipment.
A simple example would involve checking to see whether or not
a given die is balanced. Under the null hypothesis, we would assume
that the probability of a specific side coming up when we roll this
die is 1/6. To test this notion, we can roll the given die a specific
Outcome
1-spot
2-spots
3-spots
4-spots
5 -spots
6-spots
Frequency
26
31
29
33
26
35
1. Theoretically, how often would you expect each outcome to occur
if we roll this die 180 times, assuming it is a fair die?
2. Perform the appropriate hypothesis test to determine the p-value
with the null hypothesis that the die is fair. What is your conclu-
sion?
3. How much do you have to change the frequencies for various out-
comes in the above table to obtain a conclusion for the hypothesis
test of question 2 that is the opposite of the one you obtained above?
Does your conclusion switch faster if you make a big change to one
frequency and small changes to the others or if you make moderate
changes to all of the categories? (Remember that the sum of all fre-
quencies has to remain 180.)
Technology Instruction 539
ECHNOLOGY
INSTRUCTION
Chi-Square Tests
1. To perform an independence or homogeneity test on a contingency table, enter the actual
data and the expected values as matrices. To do so, select MATRX >EDIT, and use the
arrow key to select the name of your matrix. Press ENTER, and then type in the number
of rows, the number of columns, and the entries for each matrix.
Observed: [flli
Expected: [B]
Calculate Draw
Screen 11.1
Select STAT>TESTS># 2 -Test. You will need to enter the names of the Observed
and the Expected data matrices. For each entry, position the cursor, and then select
MATRX >NAMES and use the arrow keys to choose the appropriate name, and then
press ENTER. (See Screen 11.1.) After entering the matrix names, press ENTER. The
result includes the value of x 2 , the /7-value, and the degrees of freedom. (See Screen 11.2.)
X* = 16
p=3.3546263e-4
df=2
Screen 11.2
1. To perform an independence or homogeneity test on a contingency table, enter the actual
data into columns, then select Stat>Tables>Chi-square Test.
2. Enter the names of the columns containing the table, and select OK. (See Screen 11.3.)
The result includes the expected values, the degrees of freedom, the value of chi-square,
and the ;?-value.
Screen 11.3
Chi-Square Test (Table in Worksheet)
CI
C2
Columns containing the table:
cl-c2
Select
Help
OK
Cancel
540 Chapter 1 1 Chi-Square Tests
A
B
C
D
E
1
Actual
Expected
2
3
20
80
i
25
1
75
4
40
160
50
150
5
40
i
60
25
1
75
I 1
6
| J
7
p-value
=Cf3TFESTt A3 :B 5J3 3 £5)
8
[ CHTTEST(actual_range, expected_range| |
Screen 11.4
1. To perform a goodness-of-fit or independence test on a contin-
gency table, enter the actual data in a range of cells and the ex-
pected data in another range of cells with the same number of
rows and columns.
2. Type =CHITEST(actual range, expected range) and press
Enter. The result is the p-value of the test. (See Screen 11.4.)
I
TECHNOLOGY ASSIGNMENTS
TAll.l Air Quality Index (AQI) data for the city of Kitchener, Ontario (Canada), during the period
January 1, 2007 to July 14, 2009 produced the following percentage distribution.
AQI
Very good
Good
Moderate
Poor
Percentage
13.44
67.73
17.84
0.99
Source: www.airqualityontario.com.
The following table gives the AQI data for a sample of 600 readings from cities similar to Kitchener (pop-
ulation 150,000 to 250,000, metropolitan area population of 400,000 to 500,000).
AQI
Very good
Good
Moderate
Poor
Number of readings
89
393
114
4
Test at the 10% significance level whether the distribution of AQI for the sample data differs from the dis-
tribution for Kitchener, Ontario.
TA11.2 A sample of 4000 persons aged 1 8 years and older produced the following two-way classification table.
Men
Women
Single
531
357
Married
1375
1179
Widowed
55
195
Divorced
139
169
Test at the 10% significance level whether gender and marital status are dependent for all persons aged
18 years and older.
TA11.3 Two samples, one of 3000 students from urban high schools and another of 2000 students from
rural high schools, were taken. These students were asked if they have ever smoked. The following table
lists the summary of the results.
Urban
Rural
Have never smoked
1448
1228
Have smoked
1552
772
Using the 5% significance level, test the null hypothesis that the proportions of urban and rural students
who have smoked and who have never smoked are homogeneous.
Chapter
Analysis of Variance
Trying something new can be risky, and there can be uncertainty about the results. Suppose a
school district plans to test three different methods for teaching arithmetic. After teachers imple-
ment these different methods for a semester, administrators want to know if the mean scores of
students taught with these three different methods are all the same. What data will they require
and how will they test for this equality of more than two means? (See Examples 12-2 and 12-3)
12.1 The F Distribution
12.2 One-Way Analysis of
Variance
Chapter 10 described the procedures that are used to test hypotheses about the difference between
two population means using the normal and f distributions. Also described in that chapter were the
hypothesis-testing procedures for the difference between two population proportions using the normal
distribution. Then, Chapter 1 1 explained the procedures that are used to test hypotheses about the
equality of more than two population proportions using the chi-square distribution.
This chapter explains how to test the null hypothesis that the means of more than two populations
are equal. For example, suppose that teachers at a school have devised three different methods to teach
arithmetic. They want to find out if these three methods produce different mean scores. Let n h /jl 2 , and
be the mean scores of all students who will be taught by Methods I, II, and III, respectively. To test
whether or not the three teaching methods produce the same mean, we test the null hypothesis
H : /x, = ix 2 = (All three population means are equal.)
against the alternative hypothesis
H] : Not all three population means are equal.
We use the analysis of variance procedure to perform such a test of hypothesis.
Note that the analysis of variance procedure can be used to compare two population means.
However, the procedures learned in Chapter 10 are more efficient for performing tests of hypothesis
about the difference between two population means; the analysis of variance procedure, to be discussed
in this chapter, is used to compare three or more population means.
541
542 Chapter 12 Analysis of Variance
An analysis of variance test is performed using the F distribution. First, the F distribution is de-
scribed in Section 12.1 of this chapter. Then, Section 12.2 discusses the application of the one-way
analysis of variance procedure to perform tests of hypothesis.
12.1 The F Distribution
Like the chi-square distribution, the shape of a particular F distribution 1 curve depends on the
number of degrees of freedom. However, the F distribution has two numbers of degrees of free-
dom: degrees of freedom for the numerator and degrees of freedom for the denominator. These
two numbers representing two types of degrees of freedom are the parameters of the F distri-
bution. Each combination of degrees of freedom for the numerator and for the denominator gives
a different F distribution curve. The units of an F distribution are denoted by F, which assumes
only nonnegative values. Like the normal, t, and chi-square distributions, the F distribution is a
continuous distribution. The shape of an F distribution curve is skewed to the right, but the
skewness decreases as the number of degrees of freedom increases.
Definition
The F Distribution
1. The F distribution is continuous and skewed to the right.
2. The F distribution has two numbers of degrees of freedom: df for the numerator and df for
the denominator.
3. The units of an F distribution, denoted by F, are nonnegative.
For an F distribution, degrees of freedom for the numerator and degrees of freedom for the
denominator are usually written as follows:
df= (8, 14)
First number denotes the Second number denotes the
df for the numerator df for the denominator
Figure 12.1 shows three F distribution curves for three sets of degrees of freedom for the
numerator and for the denominator. In the figure, the first number gives the degrees of freedom
associated with the numerator, and the second number gives the degrees of freedom associated
with the denominator. We can observe from this figure that as the degrees of freedom increase,
the peak of the curve moves to the right; that is, the skewness decreases.
Table VII in Appendix C lists the values of F for the F distribution. To read Table VII, we
need to know three quantities: the degrees of freedom for the numerator, the degrees of freedom
for the denominator, and an area in the right tail of an F distribution curve. Note that the F dis-
tribution table (Table VII) is read only for an area in the right tail of the F distribution curve.
Figure 12.1 Three F distribution
curves.
r #=(1,3)
'The F distribution is named after Sir Ronald Fisher.
12.1 The F Distribution 543
Also note that Table VII has four parts. These four parts give the F values for areas of .01, .025,
.05, and .10, respectively, in the right tail of the F distribution curve. We can make the F distri-
bution table for other values in the right tail. Example 12-1 illustrates how to read Table VII.
■ EXAMPLE 12-1
Find the F value for 8 degrees of freedom for the numerator, 14 degrees of freedom for the
denominator, and .05 area in the right tail of the F distribution curve.
Solution To find the required value of F, we use the portion of Table VII of Appendix C
that corresponds to .05 area in the right tail of the F distribution curve. The relevant portion
of that table is shown here as Table 12.1. To find the required F value, we locate 8 in the row
Reading the F
distribution table.
Table 12.1 Obtaining the F Value From Table VII
Degrees of Freedom for the Numerator
Degrees of Freedom for
the Denominator
The F value for 8 df for the numerator,
14 d/for the denominator, and .05 area
in the right tail
for degrees of freedom for the numerator (at the top of Table VII) and 14 in the column for
degrees of freedom for the denominator (the first column on the left side in Table VII). The
entry where the column for 8 and the row for 14 intersect gives the required F value. This
value of F is 2.70, as shown in Table 12.1 and Figure 12.2. The F value obtained from this
table for a test of hypothesis is called the critical value of F.
Figure 12.2 The value of F from Table
VII for 8 df for the numerator, 14 df for
the denominator, and .05 area in the
right tail.
The required F value
EXERCISES
CONCEPTS AND PROCEDURES
12.1 Describe the main characteristics of an F distribution.
12.2 Find the critical value of F for the following.
a. df = (3, 3) and area in the right tail = .05
b. df = (3, 10) and area in the right tail = .05
c. df = (3, 30) and area in the right tail = .05
12.3 Find the critical value of F for the following.
a. df = (2, 6) and area in the right tail = .025
b. df = (6, 6) and area in the right tail = .025
c. df = (15, 6) and area in the right tail = .025
544
Chapter 12 Analysis of Variance
12.4 Determine the critical value of F for the following.
a. df = (6, 12) and area in the right tail = .01
b. df = (6, 40) and area in the right tail = .01
c. df = (6, 100) and area in the right tail = .01
12.5 Determine the critical value of F for the following.
a. df = (2, 2) and area in the right tail = .10
b. df = (8, 8) and area in the right tail = .10
c. df = (20, 20) and area in the right tail = .10
12.6 Find the critical value of F for an F distribution with df = (3, 12) and
a. area in the right tail = .05
b. area in the right tail = .10
12.7 Find the critical value of F for an F distribution with df = (11, 5) and
a. area in the right tail = .01
b. area in the right tail = .025
12.8 Find the critical value of F for an F distribution with .025 area in the right tail and
a. df= (4, 11)
b. df= (15, 3)
12.9 Find the critical value of F for an F distribution with .01 area in the right tail and
a. df= (10, 10)
b. df = (9, 25)
12.2 One-Way Analysis of Variance
As mentioned in the beginning of this chapter, the analysis of variance procedure is used to test
the null hypothesis that the means of three or more populations are the same against the alterna-
tive hypothesis that not all population means are the same. The analysis of variance procedure
can be used to compare two population means. However, the procedures learned in Chapter 10
are more efficient for performing tests of hypotheses about the difference between two popula-
tion means; the analysis of variance procedure is used to compare three or more population means.
Reconsider the example of teachers at a school who have devised three different methods to
teach arithmetic. They want to find out if these three methods produce different mean scores. Let
Hi, fi 2 , and /j, 3 be the mean scores of all students who are taught by Methods I, II, and III, respec-
tively. To test if the three teaching methods produce different means, we test the null hypothesis
H : = jx 2 = (All three population means are equal.)
against the alternative hypothesis
H\. Not all three population means are equal.
One method to test such a hypothesis is to test the three hypotheses H : ix { = fi 2 , H Q : il x =
fi 3 , and H : /x 2 = /a 3 separately using the procedure discussed in Chapter 10. Besides being time
consuming, such a procedure has other disadvantages. First, if we reject even one of these three
hypotheses, then we must reject the null hypothesis H : ptj = /jl 2 = /u, 3 . Second, combining the
Type I error probabilities for the three tests (one for each test) will give a very large Type I er-
ror probability for the test H : ix { = fi 2 = /a 3 . Hence, we should prefer a procedure that can test
the equality of three means in one test. The ANOVA, short for analysis of variance, provides
such a procedure. It is used to compare three or more population means in a single test.
Definition
ANOVA ANOVA is a procedure used to test the null hypothesis that the means of three or more
populations are all equal.
This section discusses the one-way ANOVA procedure to make tests by comparing the means
of several populations. By using a one-way ANOVA test, we analyze only one factor or variable.
For instance, in the example of testing for the equality of mean arithmetic scores of students taught
12.2 One-Way Analysis of Variance 545
by each of the three different methods, we are considering only one factor, which is the effect of
different teaching methods on the scores of students. Sometimes we may analyze the effects of
two factors. For example, if different teachers teach arithmetic using these three methods, we can
analyze the effects of teachers and teaching methods on the scores of students. This is done by
using a two-way ANOVA. The procedure under discussion in this chapter is called the analysis of
variance because the test is based on the analysis of variation in the data obtained from different
samples. The application of one-way ANOVA requires that the following assumptions hold true.
Assumptions of One-Way ANOVA The following assumptions must hold true to use one-way
ANOVA.
1. The populations from which the samples are drawn are (approximately) normally distributed.
2. The populations from which the samples are drawn have the same variance (or standard
deviation).
3. The samples drawn from different populations are random and independent.
For instance, in the example about three methods of teaching arithmetic, we first assume
that the scores of all students taught by each method are (approximately) normally distributed.
Second, the means of the distributions of scores for the three teaching methods may or may not
be the same, but all three distributions have the same variance, cr. Third, when we take sam-
ples to make an ANOVA test, these samples are drawn independently and randomly from three
different populations.
The ANOVA test is applied by calculating two estimates of the variance, cr 2 , of population
distributions: the variance between samples and the variance within samples. The variance
between samples is also called the mean square between samples or MSB. The variance within
samples is also called the mean square within samples or MSW.
The variance between samples, MSB, gives an estimate of cr based on the variation among
the means of samples taken from different populations. For the example of three teaching meth-
ods, MSB will be based on the values of the mean scores of three samples of students taught by
three different methods. If the means of all populations under consideration are equal, the means
of the respective samples will still be different, but the variation among them is expected to be
small, and, consequently, the value of MSB is expected to be small. However, if the means of
populations under consideration are not all equal, the variation among the means of respective
samples is expected to be large, and, consequently, the value of MSB is expected to be large.
The variance within samples, MSW, gives an estimate of cr based on the variation within
the data of different samples. For the example of three teaching methods, MSW will be based
on the scores of individual students included in the three samples taken from three populations.
The concept of MSW is similar to the concept of the pooled standard deviation, s p , for two sam-
ples discussed in Section 10.2 of Chapter 10.
The one-way ANOVA test is always right-tailed with the rejection region in the right tail of
the F distribution curve. The hypothesis-testing procedure using ANOVA involves the same five
steps that were used in earlier chapters. The next subsection explains how to calculate the value
of the test statistic F for an ANOVA test.
12.2.1 Calculating the Value of the Test Statistic
The value of the test statistic F for a test of hypothesis using ANOVA is given by the ratio of
two variances, the variance between samples (MSB) and the variance within samples (MSW).
Test Statistic F for a One-Way ANOVA Test The value of the test statistic F for an ANOVA test
is calculated as
Variance between samples MSB
F = or
Variance within samples MSW
The calculation of MSB and MSW is explained in Example 12-2.
546 Chapter 12 Analysis of Variance
Example 12-2 describes the calculation of MSB, MSW, and the value of the test statistic F.
Since the basic formulas are laborious to use, they are not presented here. We have used only
the short-cut formulas to make calculations in this chapter.
■ EXAMPLE 12-2
Fifteen fourth-grade students were randomly assigned to three groups to experiment with three
different methods of teaching arithmetic. At the end of the semester, the same test was given
to all 15 students. The table gives the scores of students in the three groups.
Method I
Method II
Method III
48
55
84
73
85
68
51
70
95
65
69
74
87
90
67
Calculate the value of the test statistic F. Assume that all the required assumptions mentioned
in Section 12.2 hold true.
Solution In ANOVA terminology, the three methods used to teach arithmetic are called
treatments. The table contains data on the scores of fourth-graders included in the three sam-
ples. Each sample of students is taught by a different method. Let
x = the score of a student
k = the number of different samples (or treatments)
«, = the size of sample i
Tj = the sum of the values in sample i
n = the number of values in all samples = n, + n 2 + n 3 + ■ ■ ■
Xx = the sum of the values in all samples = T l + T 2 + T 3 + ■ ■ ■
Xx 2 = the sum of the squares of the values in all samples
To calculate MSB and MSW, we first compute the between-samples sum of squares, de-
noted by SSB, and the within -samples sum of squares, denoted by SSW. The sum of SSB
and SSW is called the total sum of squares and is denoted by SST; that is,
SST = SSB + SSW
The values of SSB and SSW are calculated using the following formulas.
Calculating the value of the
test statistic F.
Between- and Within-Samples Sums of Squares The between-samples sum of squares, denoted
by SSB, is calculated as
(t\ n n \ (Sx) 2
SSB = — + — + — + ■•■ - - — -
\/t] n 2 «3 / n
The within-samples sum of squares, denoted by SSW, is calculated as
SSW = Xx 2 -
Table 12.2 lists the scores of 15 students who were taught arithmetic by each of the three
different methods; the values of 7\, T 2 , and T 3 ; and the values of n x , n 2 , and n 3 .
12.2 One-Way Analysis of Variance 547
Table 12.2
Method I
Method II
Method III
48
55
84
73
85
68
51
70
95
65
69
74
87
90
67
r, = 324
«j = 5
T 2 = 369
n 2 = 5
T 3 = 388
n 3 = 5
In Table 12.2, T x is obtained by adding the five scores of the first sample. Thus, T x = 48 +
73 + 51 + 65 + 87 = 324. Similarly, the sums of the values in the second and third samples
give T 2 = 369 and T 3 = 388, respectively. Because there are five observations in each sam-
ple, n x = n 2 = « 3 = 5. The values of Sx and n are, respectively,
Xx = T x + T 2 + T 3 = 324 + 369 + 388 = 1081
n = n x + n 2 + n 3 = 5 + 5 + 5 = 15
To calculate 2x 2 , we square all the scores included in all three samples and then add them. Thus,
£x 2 = (48) 2 + (73) 2 + (51) 2 + (65) 2 + (87) 2 + (55) 2 + (85) 2 + (70) 2
+ (69) 2 + (90) 2 + (84) 2 + (68) 2 + (95) 2 + (74) 2 + (67) 2
= 80,709
Substituting all the values in the formulas for SSB and SSW, we obtain the following val-
ues of SSB and SSW:
/(324) 2 (369) 2 (388) 2 \ (1081) 2
SSB = ( — ^— + — ^— + = 432.1333
/(324) 2 (369) 2 (388) 2 \
SSW = 80,709 - ( + + J = 2372.8000
The value of SST is obtained by adding the values of SSB and SSW. Thus,
SST = 432.1333 + 2372.8000 = 2804.9333
The variance between samples (MSB) and the variance within samples (MSW) are calculated
using the following formulas.
Calculating the Values of MSB and MSW MSB and MSW are calculated as, respectively,
SSB SSW
MSB = and MSW =
k — 1 n — k
where k — 1 and n — k are, respectively, the df for the numerator and the df for the denomina-
tor for the F distribution. Remember, k is the number of different samples.
Consequently, the variance between samples is
SSB 432.1333
MSB = = = 216.0667
k - 1 3-1
The variance within samples is
SSW 2372.8000
MSW = = = 197.7333
n-k 15-3
548 Chapter 12 Analysis of Variance
The value of the test statistic F is given by the ratio of MSB and MSW. Therefore,
MSB 216.0667 „
F = = = 1.09
MSW 197.7333
For convenience, all these calculations are often recorded in a table called the ANOVA table.
Table 12.3 gives the general form of an ANOVA table.
Table 12.3 ANOVA Table
Source of
Degrees of
Sum of
Mean
Value of the
Variation
Freedom
Squares
Square
Test Statistic
Between
k - 1
SSB
MSB
MSB
Within
n — k
SSW
MSW
MSW
Total
n - 1
SST
Substituting the values of the various quantities into Table 12.3, we write the ANOVA table
for our example as Table 12.4.
Table 12.4 ANOVA Table for Example 12-2
Source of
Variation
Degrees of
Freedom
Sum of
Squares
Mean
Square
Value of the
Test Statistic
Between
Within
2
12
432.1333
2372.8000
216.0667
197.7333
216.0667
F = = 1.09
197.7333
Total
14
2804.9333
12.2.2 One-Way ANOVA Test
Now suppose we want to test the null hypothesis that the mean scores are equal for all three
groups of fourth-graders taught by three different methods of Example 1 2-2 against the alter-
native hypothesis that the mean scores of all three groups are not equal. Note that in a one-way
ANOVA test, the null hypothesis is that the means for all populations are equal. The alternative
hypothesis is that not all population means are equal. In other words, the alternative hypothe-
sis states that at least one of the population means is different from the others. Example 12-3
demonstrates how we use the one-way ANOVA procedure to make such a test.
■ EXAMPLE 12-3
Reconsider Example 12-2 about the scores of 15 fourth-grade students who were randomly
assigned to three groups in order to experiment with three different methods of teaching arith-
metic. At the 1% significance level, can we reject the null hypothesis that the mean arithmetic
score of all fourth-grade students taught by each of these three methods is the same? Assume
that all the assumptions required to apply the one-way ANOVA procedure hold true.
Solution To make a test about the equality of the means of three populations, we follow
our standard procedure with five steps.
Step 1. State the null and alternative hypotheses.
Let fi u jj, 2 , and /jl 3 be the mean arithmetic scores of all fourth-grade students who are taught,
respectively, by Methods I, II, and III. The null and alternative hypotheses are
H : /jL t = jx 2 = (The mean scores of the three groups are all equal.)
Hi. Not all three means are equal.
Performing a one-way ANOVA
test: all samples the same size.
12.2 One-Way Analysis of Variance 549
Note that the alternative hypothesis states that at least one population mean is different from
the other two.
Step 2. Select the distribution to use.
Because we are comparing the means for three normally distributed populations, we use
the F distribution to make this test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .01. Because a one-way ANOVA test is always right-tailed, the area
in the right tail of the F distribution curve is .01, which is the rejection region in Figure 12.3.
Next we need to know the degrees of freedom for the numerator and the denominator. In
our example, the students were assigned to three different methods. As mentioned earlier, these
methods are called treatments. The number of treatments is denoted by k. The total number
of observations in all samples taken together is denoted by n. Then, the number of degrees of
freedom for the numerator is equal to k — 1 and the number of degrees of freedom for the de-
nominator is equal to n — k. In our example, there are 3 treatments (methods of teaching) and
15 total observations (total number of students) in all 3 samples. Thus,
Degrees of freedom for the numerator = A: — 1=3 — 1=2
Degrees of freedom for the denominator = n — fc=15 — 3 = 12
From Table VII of Appendix C, we find the critical value of F for 2 df for the numerator, 12
df for the denominator, and .01 area in the right tail of the F distribution curve. This value is
shown in Figure 12.3. The required value of F is 6.93.
Thus, we will fail to reject H if the calculated value of the test statistic F is less than 6.93,
and we will reject H if it is 6.93 or larger.
Step 4. Calculate the value of the test statistic.
We computed the value of the test statistic F for these data in Example 12-2. This value is
F = 1.09
Step 5. Make a decision.
Because the value of the test statistic F = 1 .09 is less than the critical value of F = 6.93, it
falls in the nonrejection region. Hence, we fail to reject the null hypothesis, and conclude that
the means of the three populations are equal. In other words, the three different methods of teach-
ing arithmetic do not seem to affect the mean scores of students. The difference in the three mean
scores in the case of our three samples occurred only because of sampling error. H
In Example 12-3, the sample sizes were the same for all treatments. Example 12-4 de-
scribes a case in which the sample sizes are not the same for all treatments.
■ EXAMPLE 12-4
From time to time, unknown to its employees, the research department at Post Bank observes
various employees for their work productivity. Recently this department wanted to check
whether the four tellers at a branch of this bank serve, on average, the same number of cus-
tomers per hour. The research manager observed each of the four tellers for a certain number
Performing a one-way
ANOVA test: all samples not
the same size.
550 Chapter 12 Analysis of Variance
of hours. The following table gives the number of customers served by the four tellers during
each of the observed hours.
Teller A
Teller B
Teller C
Teller D
19
14
11
24
21
16
14
19
26
14
21
21
24
13
13
26
18
17
16
20
13
18
At the 5% significance level, test the null hypothesis that the mean number of customers served
per hour by each of these four tellers is the same. Assume that all the assumptions required
to apply the one-way ANOVA procedure hold true.
Solution To make a test about the equality of means of four populations, we follow our
standard procedure with five steps.
Step 1. State the null and alternative hypotheses.
Let jx x , /x 2 , M3, and jx A be the mean number of customers served per hour by tellers A, B,
C, and D, respectively. The null and alternative hypotheses are, respectively,
H : ptj = fi 2 = ^3 = fa (The mean number of customers served per hour by
each of the four tellers is the same.)
Hi. Not all four population means are equal.
Step 2. Select the distribution to use.
Because we are testing for the equality of four means for four normally distributed popu-
lations, we use the F distribution to make the test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .05, which means the area in the right tail of the F distribution
curve is .05. In this example, there are 4 treatments (tellers) and 22 total observations in all
four samples. Thus,
Degrees of freedom for the numerator = ^ — 1=4—1=3
Degrees of freedom for the denominator = n — k = 22 — 4=18
The critical value of F from Table VII for 3 df for the numerator, 18 df for the denominator, and
.05 area in the right tail of the F distribution curve is 3.16. This value is shown in Figure 12.4.
Step 4. Calculate the value of the test statistic.
First we calculate SSB and SSW. Table 12.5 lists the numbers of customers served by the four
tellers during the selected hours; the values of T h T 2 , T 3 , and T 4 ; and the values of n h n 2 , n 3 , and n 4 .
The values of %x and n are, respectively,
Xx = r, + T 2 + T 3 + T 4 = 108 + 87 + 93 + 110 = 398
n = n { + n 2 + n 3 + n 4 = 5 + 6 + 6 + 5= 22
12.2 One-Way Analysis of Variance 551
Table 12.5
Teller A
Teller B
Teller C
Teller D
19
14
11
24
21
16
14
19
26
14
21
21
24
13
13
26
18
17
16
20
13
18
r, = 108
r, = 87
r 3 = 93
T 4 = 110
n\ = 5 n 2 = 6 n 3 = 6 n 4 = 5
The value of 2x 2 is calculated as follows:
2x 2 = (19) 2 + (21) 2 + (26) 2 + (24) 2 + (18) 2 + (14) 2 + (16) 2 + (14) 2
+ (13) 2 + (17) 2 + (13) 2 + (ll) 2 + (14) 2 + (21) 2 + (13) 2
+ (16) 2 + (18) 2 + (24) 2 + (19) 2 + (21) 2 + (26) 2 + (20) 2
= 7614
Substituting all the values in the formulas for SSB and SSW, we obtain the following values
of SSB and SSW:
/r 2 n n r 2 \ (£x) 2
ssb = — + — + — + — - ' — -
VH; n 2 n 3 n 4 / n
_ /(108) 2 + (87) 2 + (93) 2 + (110) 2 \ (398) 2
255.6182
V 5 6 6 5 J 22
^ , {T? Ti Ti Ti
SSW = Sx 2 - — + — + — + —
\n, n 2 n 3 n 4
/(108) 2 (87) 2 (93) 2 (110) 2 \
= 7614 - + — — + — — + = 158.2000
\ 5 6 6 5 /
Hence, the variance between samples MSB and the variance within samples MSW are,
respectively,
SSB 255.6182
MSB = = = 85.2061
k - 1 4-1
SSW 158.2000
MSW = = = 8.7889
n - k 22-4
The value of the test statistic F is given by the ratio of MSB and MSW, which is
MSB 85.2061
F = = = 9.69
MSW 8.7889
Writing the values of the various quantities in the ANOVA table, we obtain Table 12.6.
Table 12.6 ANOVA Table for Example 12-4
Source of
Variation
Degrees of
Freedom
Sum of
Squares
Mean
Square
Value of the
Test Statistic
Between
Within
3
18
255.6182
158.2000
85.2061
8.7889
F = 85 ' 2 ° 61 = 9.69
8.7889
Total
21
413.8182
552 Chapter 12 Analysis of Variance
Step 5. Make a decision.
Because the value of the test statistic F = 9.69 is greater than the critical value of F = 3.16,
it falls in the rejection region. Consequently, we reject the null hypothesis, and conclude that
the mean number of customers served per hour by each of the four tellers is not the same. In
other words, at least one of the four means is different from the other three. H
EXERCISES
CONCEPTS AND PROCEDURES
12.10 Briefly explain when a one-way ANOVA procedure is used to make a test of hypothesis.
12.11 Describe the assumptions that must hold true to apply the one-way analysis of variance procedure
to test hypotheses.
12.12 Consider the following data obtained for two samples selected at random from two populations that
are independent and normally distributed with equal variances.
Sample I Sample II
32
27
26
35
31
33
20
40
27
38
34
31
a. Calculate the means and standard deviations for these samples using the formulas from Chap-
ter 3.
b. Using the procedure learned in Section 10.2 of Chapter 10, test at the 1% significance level whether
the means of the populations from which these samples are drawn are equal.
c. Using the one-way ANOVA procedure, test at the 1% significance level whether the means of the
populations from which these samples are drawn are equal.
d. Are the conclusions reached in parts b and c the same?
12.13 Consider the following data obtained for two samples selected at random from two populations that
are independent and normally distributed with equal variances.
Sample I
Sample II
14
11
21
8
1 1
12
9
18
13
15
20
7
17
6
a. Calculate the means and standard deviations for these samples using the formulas from Chap-
ter 3.
b. Using the procedure learned in Section 10.2 of Chapter 10, test at the 5% significance level whether
the means of the populations from which these samples are drawn are equal.
c. Using the one-way ANOVA procedure, test at the 5% significance level whether the means of the
populations from which these samples are drawn are equal.
d. Are the conclusions reached in parts b and c the same?
12.2 One-Way Analysis of Variance 553
12.14 The following ANOVA table, based on information obtained for three samples selected from
three independent populations that are normally distributed with equal variances, has a few missing
values.
Source of
Variation
Degrees of
Freedom
Sum of
Squares
Mean
Square
Value of the
Test Statistic
Between
Within
2
89.3677
19.2813
F —
Total
12
a. Find the missing values and complete the ANOVA table.
b. Using a = .01, what is your conclusion for the test with the null hypothesis that the means of
the three populations are all equal against the alternative hypothesis that the means of the three
populations are not all equal?
12.15 The following ANOVA table, based on information obtained for four samples selected from four
independent populations that are normally distributed with equal variances, has a few missing values.
Source of
Variation
Degrees of
Freedom
Sum of
Squares
Mean
Square
Value of the
Test Statistic
Between
Within
15
9.2154
F — - 4.07
Total
18
a. Find the missing values and complete the ANOVA table.
b. Using a = .05, what is your conclusion for the test with the null hypothesis that the means of
the four populations are all equal against the alternative hypothesis that the means of the four
populations are not all equal?
■ APPLICATIONS
For the following exercises assume that all the assumptions required to apply the one-way ANOVA
procedure hold true.
12.16 The recommended acidity levels for sweet white wines (e.g., certain Rieslings, Port, Eiswein,
Muscat) is .70% to .85% (www.grapestompers.com/articles/measure_acidity.htm). A vintner (winemaker)
takes three random samples of Riesling from casks that are 15, 20, and 25 years old, respectively, and
measures the acidity of each sample. The sample results are given in the table below.
15 years
20 years
25 years
.8036
.8109
.7735
.8001
.8246
.7813
.8291
.8245
.8052
.8077
.8070
.8000
.8298
.8023
.8091
.8126
.8182
.7952
.8169
.8265
.7882
.8066
.8262
.7789
.8142
.8048
.7976
.8197
.7995
.7918
.8129
.8102
.7850
.8133
.7957
.7801
.8251
.8164
.7843
554 Chapter 12 Analysis of Variance
a. We are to test whether the mean acidity levels for all casks of Riesling are the same for the three
different ages. Write the null and alternative hypotheses.
b. Show the rejection and nonrejection regions on the F distribution curve for a =.025.
c. Calculate SSB, SSW, and SST.
d. What are the degrees of freedom for the numerator and the denominator?
e. Calculate the between-samples and within-samples variances.
f. What is the critical value of F for a = .025?
g. What is the calculated value of the test statistic Fl
h. Write the ANOVA table for this exercise.
i. Will you reject the null hypothesis stated in part a at a significance level of 2.5%?
12.17 A local "pick-your-own" farmer decided to grow blueberries. The farmer purchased and planted eight
plants of each of the four different varieties of highbush blueberries. The yield (in pounds) of each plant
was measured in the upcoming year to determine whether the average yields were different for at least two
of the four plant varieties. The yields of these plants of the four varieties are given in the following table.
Berkeley
5.13
5.36
5.20
5.15
4.96
5.14
5.54
5.22
Duke
5.31
4.89
5.09
5.57
5.36
4.71
5.13
5.30
Jersey
5.20
4.92
5.44
5.20
5.17
5.24
5.08
5.13
Sierra
5.08
5.30
5.43
4.99
4.89
5.30
5.35
5.26
a. We are to test whether the mean yields for all such bushes of the four varieties are the same.
Write the null and alternative hypotheses.
b. What are the degrees of freedom for the numerator and the denominator?
c. Calculate SSB, SSW, and SST.
d. Show the rejection and nonrejection regions on the F distribution curve for a = .01.
e. Calculate the between-samples and within-samples variances.
f. What is the critical value of F for a = .01?
g. What is the calculated value of the test statistic Fl
h. Write the ANOVA table for this exercise.
i. Will you reject the null hypothesis stated in part a at a significance level of 1%?
12.18 Surfer Dude swimsuit company plans to produce a new line of quick-dry swimsuits. Three textile com-
panies are competing for the company's quick-dry fabric contract. To check the fabrics of the three compa-
nies, Surfer Dude selected 10 random swatches of fabric from each company, soaked them with water, and
then measured the amount of time (in seconds) each swatch took to dry when exposed to sun and a tempera-
ture of 80°F. The following table contains the amount of time (in seconds) each of these swatches took to dry.
Company A
756
801
750
777
772
768
812
770
743
824
Company B
791
696
761
760
741
810
770
823
815
845
Company C
773
794
733
740
780
801
794
719
766
743
Using the 5% significance level, test the null hypothesis that the mean drying times for all such fabric pro-
duced by the three companies are the same.
12.19 A university employment office wants to compare the time taken by graduates with three different
majors to find their first full-time job after graduation. The following table lists the time (in days) taken
to find their first full-time job after graduation for a random sample of eight business majors, seven com-
puter science majors, and six engineering majors who graduated in May 2009.
Business
Computer Science
Engineering
136
156
126
162
113
151
135
124
163
180
128
146
148
144
178
127
147
134
176
120
144
Uses and Misuses 555
At the 5% significance level, can you conclude that the mean time taken to find their first full-time job
for all May 2009 graduates in these fields is the same?
12.20 A consumer agency wanted to find out if the mean time taken by each of three brands of medicines
to provide relief from a headache is the same. The first drug was administered to six randomly
selected patients, the second to four randomly selected patients, and the third to five randomly selected
patients. The following table gives the time (in minutes) taken by each patient to get relief from a headache
after taking the medicine.
Drug I
Drug II
Drug III
25
15
44
38
21
39
42
19
54
65
25
58
47
73
52
At the 2.5% significance level, will you conclude that the mean time taken to provide relief from a headache
is the same for each of the three drugs?
12.21 A large company buys thousands of lightbulbs every year. The company is currently considering
four brands of lightbulbs to choose from. Before the company decides which lightbulbs to buy, it wants
to investigate if the mean lifetimes of the four types of lightbulbs are the same. The company's research
department randomly selected a few bulbs of each type and tested them. The following table lists the num-
ber of hours (in thousands) that each of the bulbs in each brand lasted before being burned out.
Brand I
Brand II
Brand III
Brand IV
23
19
23
26
24
23
27
24
19
18
25
21
26
24
26
29
22
20
23
28
23
22
21
24
25
19
27
28
At the 2.5% significance level, test the null hypothesis that the mean lifetime of bulbs for each of these
four brands is the same.
USES AND MISUSES... DON'T BE LATE
Imagine that working at your company requires that staff travel fre-
quently. You want to determine if the on-time performance of any
one airline is sufficiently different from that of the remaining airlines
to warrant a preferred status with your company. The local airport
Web site publishes the scheduled and actual departure and arrival
times for the four airlines that service it. You decide to perform an
ANOVA test on the mean delay times for all airline carriers at the air-
port. The null hypothesis here is that the mean delay times for Air-
lines A, B, C, and D are all the same. The results of the ANOVA test
tell you to accept the null hypothesis: All airline carriers have the
same mean departure and arrival delay times, so that adopting a pre-
ferred status based on the on-time performance is not warranted.
When your boss tells you to redo your analysis, you should not
be surprised. The choice to study flights only at the local airport was
a good one because your company should be concerned about the
performance of an airline at the most convenient airport. A regional
airport will have a much different on-time performance profile than
a large hub airport. By mixing both arrival and departure data, how-
ever, you violated the assumption that the populations are normally
distributed. For arrival data, this assumption could be valid: The in-
fluence of high-altitude winds, local weather, and the fact that the
arrival time is an estimate in the first place result in a distribution of
arrival times around the predicted arrival times. However, departure
delays are not normally distributed. Because a flight does not leave
before its departure time but can leave after, departure delays are
skewed to the right. As the statistical methods become more sophis-
ticated, so do the assumptions regarding the characteristics of the
data. Careful attention to these assumptions is required.
556 Chapter 12 Analysis of Variance
Glossary
Analysis of variance (ANOVA) A statistical technique used to test
whether the means of three or more populations are all equal.
F distribution A continuous distribution that has two parameters:
df for the numerator and df for the denominator.
Mean square between samples or MSB A measure of the varia-
tion among the means of samples taken from different populations.
Mean square within samples or MSW A measure of the varia-
tion within the data of all samples taken from different populations.
One-way ANOVA The analysis of variance technique that ana-
lyzes one variable only.
SSB The sum of squares between samples. Also called the sum of
squares of the factor or treatment.
SST The total sum of squares given by the sum of SSB and SSW.
SSW The sum of squares within samples. Also called the sum of
squares of errors.
Supplementary Exercises
For the following exercises, assume that all the assumptions required to apply the one-way ANOVA pro-
cedure hold true.
12.22 The following table lists the numbers of violent crimes reported to police on randomly selected
days for this year. The data are taken from three large cities of about the same size.
City A
City B
City C
5
2
8
9
4
12
12
1
10
3
13
3
9
7
9
7
6
14
13
Using the 5% significance level, test the null hypothesis that the mean number of violent crimes reported
per day is the same for each of these three cities.
12.23 A music company collects data from customers who purchase CDs and MP3 downloads from them.
Each person is asked to state his or her favorite musical genre from the following list: Classic Rock, Coun-
try, Hip-Hop/Rap, Jazz, Pop, and R&B. Random samples of customers were selected from each genre.
Each customer was asked how much he or she spent (in dollars) on music purchases in the last month.
The following table gives the information (in dollars) obtained from these customers.
Classic Rock
22
35
62
17
11
59
43
Country
60
36
59
27
32
56
Hip-Hop/Rap
35
52
35
55
71
75
Jazz
13
40
27
38
31
28
22
Pop
40
17
52
59
56
24
55
R&B
24
45
36
65
58
44
51
a. At the 10% significance level, will you reject the null hypothesis that the average monthly ex-
penditures of all customers in each of the six genres are the same?
b. What is the Type I error in this case, and what is the probability of committing such an error?
Explain.
12.24 A local car dealership is interested in determining how successful their salespeople are in turning a
profit when selling a car. Specifically, they are interested in the average percentage of price markups earned
on various car sales. The following table lists the percentages of price markups for a random sample of car
sales by three salespeople at this dealership. Note that here the markups are calculated as follows. Suppose
an auto dealer pays $14,000 for a car and lists the sale price as $20,000, which gives a markup of $6000.
If the car is sold for $17,000, the markup percentage earned on this sale is 50% ($3000 is half of $6000).
Supplementary Exercises 557
Ira
23.2
26.9
27.3
34.1
30.7
31.6
43.8
Jim
19.6
41.2
60.3
34.3
52.0
23.3
39.1
44.2
Kelly
52.3
50.0
53.4
37.9
26.4
41.1
25.2
41.2
a. Test at the 5% significance level whether the average markup percentage earned on all car sales
is the same for Ira, Jim, and Kelly.
b. What is the Type I error in this case, and what is the probability of committing such an error?
Explain.
12.25 A farmer wants to test three brands of weight-gain diets for chickens to determine if the mean
weight gain for each of these brands is the same. He selected 15 chickens and randomly put each of them
on one of these three brands of diet. The following table lists the weights (in pounds) gained by these
chickens after a period of 1 month.
Brand A
Brand B
Brand C
.8
.6
1.2
1.3
1.3
.8
1.7
.6
.7
.9
.4
1.5
.6
.7
.9
a. At the 1 % significance level, can you conclude that the mean weight gain for all chickens is the
same for each of these three diets?
b. If you did not reject the null hypothesis in part a, explain the Type II error that you may have
made in this case. Note that you cannot calculate the probability of committing a Type II error
without additional information.
12.26 An ophthalmologist is interested in determining whether a golfer's type of vision (far-sightedness,
near-sightedness, no prescription) impacts how well he or she can judge distance. Random samples of
golfers from these three groups (far-sightedness, near-sightedness, no prescription) were selected, and these
golfers were blindfolded and taken to the same location on a golf course. Then each of them was asked
to estimate the distance from this location to the pin at the end of the hole. The data (in yards) given in
the following table represent how far off the estimates (let us call these errors) of these golfers were from
the actual distance. A negative value implies that the person underestimated the distance, and a positive
value implies that a person overestimated the distance.
Far-sighted
-11
-9
-8
-10
-3
-11
-8
1
-4
Near-sighted
-2
-5
-7
-8
-6
-9
2
-10
-10
No prescription
-5
1
4
3
-2
-8
Test at the 1% significance level whether the average errors in predicting distance for all golfers of the
three different vision types are the same.
12.27 A resort area has three seafood restaurants, which employ students during the summer season. The
local chamber of commerce took a random sample of five servers from each restaurant and recorded the
tips they received on a recent Friday night. The results (in dollars) of the survey are shown in the table
below. Assume that the Friday night for which the data were collected is typical of all Friday nights of
the summer season.
Barzini's
Hwang's
Jack's
97
67
93
114
85
102
105
92
98
85
78
80
120
90
91
558 Chapter 12 Analysis of Variance
a. Would a student seeking a server's job at one of these three restaurants conclude that the mean
tips on a Friday night are the same for all three restaurants? Use the 5% level of significance.
b. What will your decision be in part a if the probability of making a Type I error is zero? Explain.
12.28 A student who has a 9 a.m. class on Monday, Wednesday, and Friday mornings wants to know if
the mean time taken by students to find parking spaces just before 9 a.m. is the same for each of these
three days of the week. He randomly selects five weeks and records the time taken to find a parking
space on Monday, Wednesday, and Friday of each of these five weeks. These times (in minutes) are given
in the following table. Assume that this student is representative of all students who need to find a park-
ing space just before 9 a.m. on these three days.
Monday
Wednesday
Friday
6
9
3
12
12
2
15
5
10
14
14
7
10
13
5
At the 5% significance level, test the null hypothesis that the mean time taken to find a parking space just
before 9 a.m. on Monday, Wednesday, and Friday is the same for all students.
Advanced Exercises
12.29 A billiards parlor in a small town is open just 4 days per week — Thursday through Sunday. Rev-
enues vary considerably from day to day and week to week, so the owner is not sure whether some days
of the week are more profitable than others. He takes random samples of 5 Thursdays, 5 Fridays, 5 Sat-
urdays, and 5 Sundays from last year's records and lists the revenues for these 20 days. His bookkeeper
finds the average revenue for each of the four samples, and then calculates Sx 2 . The results are shown in
the following table. The value of the Sx 2 came out to be 2,890,000.
Day
Mean Revenue ($)
Sample Size
Thursday
295
5
Friday
380
5
Saturday
405
5
Sunday
345
5
Assume that the revenues for each day of the week are normally distributed and that the standard devia-
tions are equal for all four populations. At the 1% level of significance, can you conclude that the mean
revenue is the same for each of the four days of the week?
12.30 Suppose that you are a reporter for a newspaper whose editor has asked you to compare the hourly
wages of carpenters, plumbers, electricians, and masons in your city. Since many of these workers are not
union members, the wages vary considerably among individuals in the same trade.
a. What data should you gather, and how would you collect them? What statistics would you pres-
ent in your article, and how would you calculate them? Assume that your newspaper is not in-
tended for technical readers.
b. Suppose that you must submit your findings to a technical journal that requires statistical analy-
sis of your data. If you want to determine whether or not the mean hourly wages are the same
for all four trades, briefly describe how you would analyze the data. Assume that hourly wages
in each trade are normally distributed and that the four variances are equal.
12.31 The editor of an automotive magazine has asked you to compare the mean gas mileages of city
driving for three makes of compact cars. The editor has made available to you one car of each of the three
makes, three drivers, and a budget sufficient to buy gas and pay the drivers for approximately 500 miles
of city driving for each car.
a. Explain how you would conduct an experiment and gather the data for a magazine article com-
paring the gas mileage.
b. Suppose that you wish to test the null hypothesis that the mean gas mileages of city driving are
the same for all three makes. Outline the procedure for using your data to conduct this test. As-
sume that the assumptions for applying analysis of variance are satisfied.
Self-Review Test 559
12.32 Do rock music CDs and country music CDs give the consumers the same amount of music listen-
ing time? A sample of 12 randomly selected single rock music CDs and a sample of 14 randomly selected
single country music CDs have the following total lengths (in minutes).
Rock Music
Country Music
A'x n
A^ 1
AA 1
44. J
AC\ 1
4U.Z
UJ.O
A1 C
4Z.0
11 C
11 n
<ZA 1
D4.Z
11 ^
^ 1 1
J L.J
11 1
51.1
ut.o
JU.O
36.1
34.6
33.9
33.4
51.7
36.5
36.5
43.3
59.7
31.7
44.0
42.7
Assume that the two populations are normally distributed with equal standard deviations.
a. Compute the value of the test statistic / for testing the null hypothesis that the mean lengths of
the rock and country music single CDs are the same against the alternative hypothesis that these
mean lengths are not the same. Use the value of this t statistic to compute the (approximate)
p-value.
b. Compute the value of the (one-way ANOVA) test statistic F for performing the test of equality
of the mean lengths of the rock and country music single CDs and use it to find the (approxi-
mate) p-value.
c. How do the test statistics in parts a and b compare? How do the p-values computed in parts a
and b compare? Do you think that this is a coincidence, or will this always happen?
12.33 Suppose you are performing a one-way ANOVA with only the information given in the following table.
Source of Variation
Degrees of Freedom
Sum of Squares
Between
4
200
Within
45
3547
a. Suppose the sample sizes for all groups are equal. How many groups are there? What are the
group sample sizes?
b. The p-value for the test of the equality of the means of all populations is calculated to be .6406.
Suppose you plan to increase the sample sizes for all groups but keep them all equal. However, when
you do this, the sum of squares within samples and the sum of squares between samples (magically)
remain the same. What are the smallest sample sizes for groups that would make this result signif-
icant at the 5% significance level?
Self-Review Test
1. The F distribution is
a. continuous b. discrete c. neither
2. The F distribution is always
a. symmetric b. skewed to the right c. skewed to the left
3. The units of the F distribution, denoted by F, are always
a. nonpositive b. positive c. nonnegative
560
Chapter 12 Analysis of Variance
4. The one-way ANOVA test analyzes only one
a. variable b. population c. sample
5. The one-way ANOVA test is always
a. right-tailed b. left-tailed c. two-tailed
6. For a one-way ANOVA with k treatments and n observations in all samples taken together, the degrees
of freedom for the numerator are
a. k — 1 b. n — k c. n — 1
7. For a one-way ANOVA with k treatments and n observations in all samples taken together, the degrees
of freedom for the denominator are
a. k — 1 b. n — k c. n — 1
8. The ANOVA test can be applied to compare
a. three or more population means
b. more than four population means only
c. more than three population means only
9. Briefly describe the assumptions that must hold true to apply the one-way ANOVA procedure as men-
tioned in this chapter.
10. A small college town has four pizza parlors that make deliveries. A student doing a research paper
for her business management class decides to compare how promptly the four parlors deliver. On six ran-
domly chosen nights, she orders a large pepperoni pizza from each establishment, then records the elapsed
time until the pizza is delivered to her apartment. Assume that her apartment is approximately the same
distance from the four pizza parlors. The following table shows the times (in minutes) for these deliver-
ies. Assume that all the assumptions required to apply the one-way ANOVA procedure hold true.
Tony's
Luigi's
Angelo's
Kowalski's
20.0
22.1
22.3
23.9
24.0
27.0
26.0
24.1
18.3
20.2
24.0
25.8
22.0
32.0
30.1
29.0
20.8
26.0
28.0
25.0
19.0
24.8
25.8
24.2
a. Using the 5% significance level, test the null hypothesis that the mean delivery time is the same
for each of the four pizza parlors.
b. Is it a Type I error or a Type II error that may have been committed in part a? Explain.
Mini-Projects
■ MINI-PROJECT 12-1
Are some days of the week busier than others on the New York Stock Exchange (NYSE)? Record the
number of shares traded on the NYSE each day for a period of 6 weeks (round the number of shares to
the nearest million). You will have five samples — first for shares traded on 6 Mondays, second for shares
traded on 6 Tuesdays, and so forth. Assume that these days make up random samples for the respective
populations. Further assume that each of the five populations from which these five samples are taken fol-
lows a normal distribution with the same variance. Test if the mean number of shares traded is the same
for each of the five populations. Use a 1% significance level.
■ MINI-PROJECT 12-2
Pick at least 30 students at random and divide them randomly into three groups (A, B, and C) of approxi-
mately equal size. Take the students one by one, ring a bell, and 17 seconds later ring another bell. Then ask
the students to estimate the elapsed time between the first and second rings. For group A, tell each student be-
fore the experiment starts that people tend to underestimate the elapsed time. Tell each student in group B that
people tend to overestimate the time. Do not make any such statement to the students in group C. Record the
estimates for all students, and then conduct an appropriate hypothesis test to see if the mean estimates of
elapsed time are all equal for the populations represented by these groups. Use the 5% level of significance,
and assume that the three populations of elapsed time are normally distributed with equal standard deviations.
Decide for Yourself 561
■ MINI-PROJECT 12-3
Obtain a Wiffle™ ball, a plastic golf ball with dimples and no holes, and a plastic golf ball with holes in-
stead of dimples. Throw each ball 20 times and measure the distances. Perform a hypothesis test to deter-
mine if the average distance is the same for each type of ball. Use a significance level of 5%.
■ MINI-PROJECT 12-4
Using Data Set III (NBA Data) that is on the Web site of this text, take a random sample of 15 guards,
15 forwards, and 15 centers.
a. Perform an analysis of variance to test the hypothesis that the mean salaries of guards, forwards,
and centers are all the same versus the alternative that at least two of the positions have different
average salaries. Use a 5% significance level.
b. Create a stacked dotplot of the data. (See Chapter 2 in case you need a review of how to make a
stacked dotplot.) Use this dotplot to explain the conclusion that you reached in part a.
c. Using the stacked dotplot that you made in part b, discuss whether the underlying conditions are
reasonable for this analysis. Specifically, discuss whether it seems reasonable to assume that the
salaries are normally distributed and that the salary variances are equal for the three positions.
DECIDE FOR YOURSELF
Deciding About Heights of Basketball
Players and Where They Come From
One-way ANOVA has given you a method/procedure to compare
three or more means obtained from independent samples to make a
decision about the corresponding population means. If you fail to
reject the null hypothesis, you conclude that the assumption that the
means of all populations under consideration are equal is a reasonable
assumption. However, if you reject the null hypothesis, you conclude
that at least two of the population means are different. Of course,
there is still a glaring piece of information that you need in the latter
case. If at least two means are different, which ones are different?
To determine which two means are different requires what is
called a pairwise comparison procedure. This type of procedure
compares each pair of means to determine whether or not they are
equal. There are many such procedures available that can be used to
make these pairwise comparisons. Some of these procedures are the
Tukey HSD, Bonferroni, Scheffe, and Tamhane T2. To select the
method that should be used depends on conditions such as whether
or not the sample sizes are equal and whether or not using a pooled
variance is reasonable.
There are a few informal (or ad hoc) methods that can be used
to have an idea about what might happen with the pairwise compar-
isons. It is very important to note that the results from these proce-
dures depend on how well the data meet the assumptions of an
ANOVA, so these methods are not a substitute for a formal statisti-
cal process. These informal methods are simply graphical methods
that can help you understand what is going on in a data set.
The accompanying figure gives a side-by-side plot of 95% con-
fidence intervals for the mean heights of NBA players who entered
the league from one of three sources — colleges, foreign countries, or
directly from high schools. The horizontal lines in these intervals
represent the ends of the intervals, while the circles identify the val-
ues of the sample means for the three groups. These confidence
intervals are based on random samples of 15 players selected from
each of the three groups. It is important to note that the
Interval Plot of Ht vs Source
95% a far Hie Mui
95.0-1
77,5
1 1 1
college Foreign Hicfn School
Source
condition — < .05 is not met for the high school and foreign players,
but we will not address this issue at this time. Answer the following
questions.
1. From the graph, we observe that players from one source seem
to be significantly taller or shorter, on average, than players from
the other two sources. Identify the source for which this is the case,
the specific difference (taller or shorter), and what characteristic
of the graph led you to make this conclusion.
2. The confidence interval for the NBA players who entered the
league from high school is much narrower than the confidence inter-
vals for the players who entered the league from the other two
sources, yet the standard deviations of heights for all players in each
of these three groups are relatively close. What does this tell you
about the affect of random sampling on summary statistics?
3. What does your conclusion in question 1 imply about the types of
players (centers, forwards, or guards) who come from these three
sources?
562 Chapter 12 Analysis of Variance
ECHNOLOGY
INSTRUCTION
Screen 12.1
Analysis of Variance
1. To perform a one-way analysis of variance on a collection of samples, store the sample
data in lists.
2. Select STAT >TESTS >ANOVA(.
3. Enter the names of the lists, separated by commas, and then type
a right parenthesis. Press ENTER. (See Screen 12.1.)
4. The results include the F statistic for performing the test, as well
as the p-value. (See Screen 12.2.)
One-way ANOUfl
F=4. 890046614
p=. 0279723575
Factor-
df=2
SS=1 188. 93333
I MS=594. 466667
Screen 12.2
1. To perform a one-way analysis of variance on a collection of samples, enter the data for
samples into columns.
2. Select Stat>ANOVA>One-way (Unstacked).
3. Enter the names of the columns and select OK. (See Screen 12.3.)
4. The results include the components of the ANOVA, including the ;?-value, as well as the
95% confidence interval for each population mean using a pooled estimate of the variance.
One-Way Analysis of Variance
Responses [in separate columns):
cl-c3
I Store residuals
Store fits
Confidence level: 1 9 E . □
Selecl
Comparisons.., Graphs
Help
OK
Cancel
Screen 12.3
1. Click the Data tab. Click the Data Analysis button within the Analysis group. From the
Data Analysis window that appears, select Anova: Single Factor.
2. Enter the location of the data in the Input Range box. Click the button to identify
whether the data for each sample are given in columns or rows. Enter the significance
level, as a decimal, in the Alpha box. If your data have labels in the top row (or in the left
column), click the Labels box. Choose how you wish the output to appear. (See Screen
12.4.) Click OK.
Technology Assignments 563
H A
c
1
IS
19
33
i
1
2
21
16
29
3
15
22
27
4
15
25
30
1
5
13
27"
30
6
31
25
7
8
9
10
11
12
13
14
D
Anova: Single Factor
Input
Input Range:
Grouped By:
I I Labels in First Row
Alpha:
0,05
Output options
O Output Range:
New Worksheet Ply:
O New Workbook
© Columns
O Rows
mm
m
OK
Cancel
Help
Screen 12.4
3. The output contains the summary statistics for each group, as well as the ANOVA table.
In addition to all of the standard items, the ANOVA table contains the critical value of F
for the given significance level and degrees of freedom. (See Screen 12.5.)
A
B
C
D
E
F
G
1
Anova: Single Factor
2
3
SUMMARY
4
Groups
Count
Sum
Average
Variance
5
Column 1
5
S2
16.4
9.S
6
Column 2
6
140
23.33333
23.36667
7
Column 3
5
174
29
7.6
B
3
10
ANOVA
11
Source of Variation
SS
MS
F
P- value
Fcrit
12
Between Groups
432.9961
2
216.498
13.37981
0.000564
3.73SS92
13
Within Groups
226.5333
14
16.1S095
14
15
Total
659.5294
16
Screen 12.5
TECHNOLOGY ASSIGNMENTS
TA12.1
TA12.2
TA12.3
Solve Exercise 12.16.
Solve Exercise 12.17.
Solve Exercise 12.24.
Simple Linear Regression
13.T Simple Linear
Regression Model
13.2 Simple Linear
Regression Analysis
Case Study 13-1 Regression
of Heights and Weights
of NBA Players
13.3 Standard Deviation of
Random Errors
1 3.4 Coefficient of
Determination
13.5 Inferences About B
1 3.6 Linear Correlation
1 3.7 Regression Analysis: A
Complete Example
1 3.8 Using the Regression
Model
13.9 Cautions in Using
Regression
Are the heights and weights of persons related? Does a person's weight depend on his/her height?
If yes, what is the change in the weight of a person, on average, for every one inch increase in
height? What is this rate of change for the National Basketball Association players? (See Case
Study 13-1).
This chapter considers the relationship between two variables in two ways: (1) by using regression
analysis and (2) by computing the correlation coefficient. By using the regression model, we can eval-
uate the magnitude of change in one variable due to a certain change in another variable. For example,
an economist can estimate the amount of change in food expenditure due to a certain change in the
income of a household by using the regression model. A sociologist may want to estimate the
increase in the crime rate due to a particular increase in the unemployment rate. Besides answering
these questions, a regression model also helps predict the value of one variable for a given value of
another variable. For example, by using the regression line, we can predict the (approximate) food
expenditure of a household with a given income.
The correlation coefficient, on the other hand, simply tells us how strongly two variables are
related. It does not provide any information about the size of the change in one variable as a result
of a certain change in the other variable. For example, the correlation coefficient tells us how strongly
income and food expenditure or crime rate and unemployment rate are related.
564
13.1 Simple Linear Regression Model
565
13.1 Simple Linear Regression Model
Only simple linear regression will be discussed in this chapter. 1 In the next two subsections the
meaning of the words simple and linear as used in simple linear regression is explained.
13.1.1 Simple Regression
Let us return to the example of an economist investigating the relationship between food ex-
penditure and income. What factors or variables does a household consider when deciding how
much money it should spend on food every week or every month? Certainly, income of the
household is one factor. However, many other variables also affect food expenditure. For in-
stance, the assets owned by the household, the size of the household, the preferences and tastes
of household members, and any special dietary needs of household members are some of the
variables that influence a household's decision about food expenditure. These variables are called
independent or explanatory variables because they all vary independently, and they explain
the variation in food expenditures among different households. In other words, these variables
explain why different households spend different amounts of money on food. Food expenditure
is called the dependent variable because it depends on the independent variables. Studying the
effect of two or more independent variables on a dependent variable using regression analysis
is called multiple regression. However, if we choose only one (usually the most important) in-
dependent variable and study the effect of that single variable on a dependent variable, it is
called a simple regression. Thus, a simple regression includes only two variables: one inde-
pendent and one dependent. Note that whether it is a simple or a multiple regression analysis,
it always includes one and only one dependent variable. It is the number of independent vari-
ables that changes in simple and multiple regressions.
Definition
Simple Regression A regression model is a mathematical equation that describes the
relationship between two or more variables. A simple regression model includes only two
variables: one independent and one dependent. The dependent variable is the one being
explained, and the independent variable is the one used to explain the variation in the dependent
variable.
13.1.2 Linear Regression
The relationship between two variables in a regression analysis is expressed by a mathematical
equation called a regression equation or model. A regression equation, when plotted, may as-
sume one of many possible shapes, including a straight line. A regression equation that gives a
straight-line relationship between two variables is called a linear regression model; otherwise,
the model is called a nonlinear regression model. In this chapter, only linear regression models
are studied.
Definition
Linear Regression A (simple) regression model that gives a straight-line relationship between
two variables is called a linear regression model.
'The term regression was first used by Sir Francis Galton (1822-1911), who studied the relationship between the
heights of children and the heights of their parents.
566 Chapter 13 Simple Linear Regression
The two diagrams in Figure 13.1 show a linear and a nonlinear relationship between the
dependent variable food expenditure and the independent variable income. A linear relationship
between income and food expenditure, shown in Figure 13.1a, indicates that as income increases,
the food expenditure always increases at a constant rate. A nonlinear relationship between in-
come and food expenditure, as depicted in Figure 13. lb, shows that as income increases, the
food expenditure increases, although, after a point, the rate of increase in food expenditure is
lower for every subsequent increase in income.
Figure 13.1 Relationship between food expenditure
and income, (a) Linear relationship, (b) Nonlinear
relationship.
Income Income
(a) (b)
The equation of a linear relationship between two variables x and y is written as
y = a + bx
Each set of values of a and b gives a different straight line. For instance, when a = 50 and
b = 5, this equation becomes
y = 50 + 5x
To plot a straight line, we need to know two points that lie on that line. We can find two
points on a line by assigning any two values to x and then calculating the corresponding values
of y. For the equation y = 50 + 5x:
1. When x = 0, then y = 50 + 5(0) = 50.
2. When x = 10, then y = 50 + 5(10) = 100.
These two points are plotted in Figure 13.2. By joining these two points, we obtain the line rep-
resenting the equation y = 50 + 5x.
Figure 13.2 Plotting a linear equation. yy
150
100
50
5 10 15 x
Note that in Figure 13.2 the line intersects the y (vertical) axis at 50. Consequently, 50 is
called the _y-intercept. The y-intercept is given by the constant term in the equation. It is the
value of y when x is zero.
In the equation y = 50 + 5x, 5 is called the coefficient of x or the slope of the line. It
gives the amount of change in y due to a change of one unit in x. For example:
Ifx = 10, theny = 50 + 5(10) = 100.
Ifx = 11, theny = 50 + 5(11) = 105.
Hence, as x increases by 1 unit (from 10 to 11), y increases by 5 units (from 100 to 105). This
is true for any value of x. Such changes in x and y are shown in Figure 13.3.
13.2 Simple Linear Regression Analysis
567
Figure 1 3.3 y-intercept and slope of a line.
y-intercept
x
In general, when an equation is written in the form
y = a + bx
a gives the y-intercept and b represents the slope of the line. In other words, a represents the
point where the line intersects the y-axis, and b gives the amount of change in y due to a change
of one unit in x. Note that b is also called the coefficient of x.
13.2 Simple Linear Regression Analysis
In a regression model, the independent variable is usually denoted by x, and the dependent vari-
able is usually denoted by v. The x variable, with its coefficient, is written on the right side of
the = sign, whereas the y variable is written on the left side of the = sign. The v-intercept and
the slope, which we earlier denoted by a and b, respectively, can be represented by any of the
many commonly used symbols. Let us denote the y-intercept (which is also called the constant
term) by A, and the slope (or the coefficient of the x variable) by B. Then, our simple linear re-
gression model is written as
Constant term or y-intercept , _ Slope
y = A + Bx (1)
"i i
Dependent variable Independent variable
In model (1), A gives the value of y for x = 0, and B gives the change in y due to a change of
one unit in x.
Model (1) is called a deterministic model. It gives an exact relationship between x and y.
This model simply states that y is determined exactly by x, and for a given value of x there is
one and only one (unique) value of y.
However, in many cases the relationship between variables is not exact. For instance, if y
is food expenditure and x is income, then model (1) would state that food expenditure is deter-
mined by income only and that all households with the same income spend the same amount
on food. As mentioned earlier, however, food expenditure is determined by many variables, only
one of which is included in model (1). In reality, different households with the same income
spend different amounts of money on food because of the differences in the sizes of the house-
hold, the assets they own, and their preferences and tastes. Hence, to take these variables into
consideration and to make our model complete, we add another term to the right side of model (1).
This term is called the random error term. It is denoted by e (Greek letter epsilon). The com-
plete regression model is written as
y = A + Bx + e (2)
f
Random error term
The regression model (2) is called a probabilistic model or a statistical relationship.
568 Chapter 13 Simple Linear Regression
Definition
Equation of a Regression Model In the regression model y = A + Bx + e, A is called the
y-intercept or constant term, B is the slope, and e is the random error term. The dependent and
independent variables are y and x, respectively.
The random error term e is included in the model to represent the following two phenomena:
1. Missing or omitted variables. As mentioned earlier, food expenditure is affected by many
variables other than income. The random error term e is included to capture the effect of
all those missing or omitted variables that have not been included in the model.
2. Random variation. Human behavior is unpredictable. For example, a household may have
many parties during one month and spend more than usual on food during that month.
The same household may spend less than usual during another month because it spent
quite a bit of money to buy furniture. The variation in food expenditure for such reasons
may be called random variation.
In model (2), A and B are the population parameters. The regression line obtained for
model (2) by using the population data is called the population regression line. The values of
A and B in the population regression line are called the true values of the y-intercept and
slope, respectively.
However, population data are difficult to obtain. As a result, we almost always use sample
data to estimate model (2). The values of the y-intercept and slope calculated from sample data
on x and y are called the estimated values of A and B and are denoted by a and b, respec-
tively. Using a and b, we write the estimated regression model as
y = a + bx (3)
where y (read as y hat) is the estimated or predicted value of y for a given value of x. Equa-
tion (3) is called the estimated regression model; it gives the regression of y on x.
Definition
Estimates of A and B In the model y = a + bx, a and b, which are calculated using sample
data, are called the estimates of A and B, respectively.
13.2.1 Scatter Diagram
Suppose we take a sample of seven households from a small city and collect information on
their incomes and food expenditures for the last month. The information obtained (in hundreds
of dollars) is given in Table 13.1.
Table 13.1 Incomes and Food Expendi-
tures of Seven Households
Income
Food Expenditure
55
14
83
24
38
13
61
16
33
9
49
15
67
17
13.2 Simple Linear Regression Analysis 569
In Table 13.1, we have a pair of observations for each of the seven households. Each pair
consists of one observation on income and a second on food expenditure. For example, the first
household's income for the last month was $5500 and its food expenditure was $1400. By plot-
ting all seven pairs of values, we obtain a scatter diagram or scatterplot. Figure 13.4 gives
the scatter diagram for the data of Table 13.1. Each dot in this diagram represents one house-
hold. A scatter diagram is helpful in detecting a relationship between two variables. For exam-
ple, by looking at the scatter diagram of Figure 13.4, we can observe that there exists a strong
linear relationship between food expenditure and income. If a straight line is drawn through the
points, the points will be scattered closely around the line.
§24 -
ll8-
Q.
S 12 -
"D
£ 6-
Seventh —
household
First
household
Figure 13.4 Scatter diagram.
20
40
60
Income
80
100
Definition
Scatter Diagram A plot of paired observations is called a scatter diagram.
As shown in Figure 13.5, a large number of straight lines can be drawn through the scatter
diagram of Figure 13.4. Each of these lines will give different values for a and b of model (3).
In regression analysis, we try to find a line that best fits the points in the scatter diagram.
Such a line provides the best possible description of the relationship between the dependent
and independent variables. The least squares method, discussed in the next section, gives
such a line. The line obtained by using the least squares method is called the least squares
regression line.
13.2.2 Least Squares Line
The value of y obtained for a member from the survey is called the observed or actual value
of y. As mentioned earlier in this section, the value of y, denoted by y, obtained for a given x
by using the regression line is called the predicted value of y. The random error e denotes the
difference between the actual value of y and the predicted value of y for population data. For
example, for a given household, e is the difference between what this household actually spent
on food during the last month and what is predicted using the population regression line. The
€ is also called the residual because it measures the surplus (positive or negative) of actual food
expenditure over what is predicted by using the regression model. If we estimate model (2) by
570 Chapter 13 Simple Linear Regression
using sample data, the difference between the actual y and the predicted y based on this esti-
mation cannot be denoted by e. The random error for the sample regression model is denoted
by e. Thus, e is an estimator of e. If we estimate model (2) using sample data, then the value
of e is given by
e = Actual food expenditure — Predicted food expenditure = y — y
In Figure 13.6, e is the vertical distance between the actual position of a household and the
point on the regression line. Note that in such a diagram, we always measure the dependent
variable on the vertical axis and the independent variable on the horizontal axis.
The value of an error is positive if the point that gives the actual food expenditure is above
the regression line and negative if it is below the regression line. The sum of these errors is
always zero. In other words, the sum of the actual food expenditures for seven households
included in the sample will be the same as the sum of the food expenditures predicted from the
regression model. Thus,
Se = X(y - 5>) =
Hence, to find the line that best fits the scatter of points, we cannot minimize the sum of errors.
Instead, we minimize the error sum of squares, denoted by SSE, which is obtained by adding
the squares of errors. Thus,
SSE = Xe 2 = X(y - yf
The least squares method gives the values of a and b for model (3) such that the sum of squared
errors (SSE) is minimum.
Error Sum of Squares (SSE) The error sum of squares, denoted by SSE, is
SSE = 2e 2 = %{y - yf
The values of a and b that give the minimum SSE are called the least squares estimates of A
and B, and the regression line obtained with these estimates is called the least squares line.
The Least Squares Line For the least squares regression line y = a + bx,
SSxy
b = and a = y — bx
ss,, y
(2x)(Xy) (Sx) 2
where SS,,, = %xy and SS„ = Xx 2
and SS stands for "sum of squares." The least squares regression line y = a + bx is also called
the regression of y on x.
13.2 Simple Linear Regression Analysis 571
The least squares values of a and b are computed using the formulas just given. 2 These for-
mulas are for estimating a sample regression line. Suppose we have access to a population data
set. We can find the population regression line by using the same formulas with a little adap-
tation. If we have access to population data, we replace a by A, b by B, and n by N in these for-
mulas, and use the values of Xx, %y, Xxy, and Xx 2 calculated for population data to make the
required computations. The population regression line is written as
fLy\ x = A + Bx
where fiyi x is read as "the mean value of y for a given x." When plotted on a graph, the points
on this population regression line give the average values of y for the corresponding values of x.
These average values of y are denoted by jx y \ x .
Example 13-1 illustrates how to estimate a regression line for sample data.
EXAMPLE 13-1
Find the least squares regression line for the data on incomes and food expenditures on the
seven households given in Table 13.1. Use income as an independent variable and food ex-
penditure as a dependent variable.
Solution We are to find the values of a and b for the regression model y = a + bx. Table
13.2 shows the calculations required for the computation of a and b. We denote the inde-
pendent variable (income) by x and the dependent variable (food expenditure) by y, both in
hundreds of dollars.
Table 13.2
Estimating the least squares
regression line.
Income
Food Expenditure
X
y
xy
x 2
55
14
770
3025
83
24
mi
6889
38
13
494
1444
61
16
976
3721
33
9
297
1089
49
15
735
2401
67
17
1139
4489
Xx = 386
Xy = 108
Xxy = 6403
Xx 2 = 23,058
The following steps are performed to compute a and b.
Step 1. Compute Xx, Xy, x, and y.
Xx = 386, Xy = 108
x = Xx/n = 386/7 = 55.1429
y = ty/n = 108/7 = 15.4286
Step 2. Compute Xxy and Xx 2 .
To calculate Xxy, we multiply the corresponding values of x and y. Then, we sum all the
products. The products of x and y are recorded in the third column of Table 13.2. To compute
"The values of SS n , and SS„ can also be obtained by using the following basic formulas:
SS tv = %{x - x)(y - y) and SS„ = Z(x - xf
However, these formulas take longer to make calculations.
572 Chapter 13 Simple Linear Regression
Xx , we square each of the x values and then add them. The squared values of x are listed in
the fourth column of Table 13.2. From these calculations,
Xxy = 6403 and %x 2 = 23,058
Step 3. Compute SS„ and SS, t :
(Zx)(Xy) (386)(108)
SS„ = Xxy — = 6403 = 447.5714
n 1
, (Xx) 2 (386) 2
SS XV = Xx 2 = 23,058 — = 1772.8571
Step 4. Compute a and b:
n ' 7
SS„ 447.5714
.2525
SS.« 1772.8571
a = y - bx = 15.4286 - (.2525)(55.1429) = 1.5050
Thus, our estimated regression model y = a + bx is
y = 1.5050 + .2525*
This regression line is called the least squares regression line. It gives the regression of food
expenditure on income.
Note that we have rounded all calculations to four decimal places. We can round the values
of a and b in the regression equation to two decimal places, but we do not do this here because
we will use this regression equation for prediction and estimation purposes later. H
Using this estimated regression model, we can find the predicted value of y for any specific
value of x. For instance, suppose we randomly select a household whose monthly income is
$6100, so that x = 61 (recall that x denotes income in hundreds of dollars). The predicted value
of food expenditure for this household is
y = 1.5050 + (.2525)(61) = $16.9075 hundred = $1690.75
In other words, based on our regression line, we predict that a household with a monthly income
of $6100 is expected to spend $1690.75 per month on food. This value of y can also be inter-
preted as a point estimator of the mean value of y for x = 61. Thus, we can state that, on aver-
age, all households with a monthly income of $6100 spend about $1690.75 per month on food.
In our data on seven households, there is one household whose income is $6100. The ac-
tual food expenditure for that household is $1600 (see Table 13.1). The difference between the
actual and predicted values gives the error of prediction. Thus, the error of prediction for this
household, which is shown in Figure 13.7, is
e=y-y = 16- 16.9075 = -$.9075 hundred = -$90.75
Figure 13.7 Error of prediction.
20 40 60 80 100
Income
Therefore, the error of prediction is —$90.75. The negative error indicates that the predicted
value of y is greater than the actual value of v. Thus, if we use the regression model, this house-
hold's food expenditure is overestimated by $90.75.
13.2 Simple Linear Regression Analysis 573
13.2.3 Interpretation of a and b
How do we interpret a = 1.5050 and b = .2525 obtained in Example 13-1 for the regression of
food expenditure on income? A brief explanation of the y-intercept and the slope of a regression
line was given in Section 13.1.2. Below we explain the meaning of a and b in more detail.
Interpretation of a
Consider a household with zero income. Using the estimated regression line obtained in Exam-
ple 13—1, we get the predicted value of y for x = as
y = 1.5050 + .2525(0) = $1.5050 hundred = $150.50
Thus, we can state that a household with no income is expected to spend $150.50 per month
on food. Alternatively, we can also state that the point estimate of the average monthly food ex-
penditure for all households with zero income is $150.50. Note that here we have used y as a
point estimate of fi y i x . Thus, a = 150.50 gives the predicted or mean value of y for x = based
on the regression model estimated for the sample data.
However, we should be very careful when making this interpretation of a. In our sample of
seven households, the incomes vary from a minimum of $3300 to a maximum of $8300. (Note that
in Table 13.1, the minimum value of x is 33 and the maximum value is 83.) Hence, our regression
line is valid only for the values of x between 33 and 83. If we predict y for a value of x outside this
range, the prediction usually will not hold true. Thus, since x = is outside the range of house-
hold incomes that we have in the sample data, the prediction that a household with zero income
spends $150.50 per month on food does not carry much credibility. The same is true if we try to
predict y for an income greater than $8300, which is the maximum value of x in Table 13.1.
Interpretation of b
The value of b in a regression model gives the change in y (dependent variable) due to a change
of one unit in x (independent variable). For example, by using the regression equation obtained
in Example 13-1, we see:
Whenx = 50, y = 1.5050 + .2525(50) = 14.1300
Whenx = 51, y = 1.5050 + .2525(51) = 14.3825
Hence, when x increased by one unit, from 50 to 5 1, y increased by 14.3825 — 14. 1300 = .2525,
which is the value of b. Because our unit of measurement is hundreds of dollars, we can state
that, on average, a $100 increase in income will result in a $25.25 increase in food expenditure.
We can also state that, on average, a $1 increase in income of a household will increase the
food expenditure by $.2525. Note the phrase "on average" in these statements. The regression
line is seen as a measure of the mean value of y for a given value of x. If one household's in-
come is increased by $100, that household's food expenditure may or may not increase by
$25.25. However, if the incomes of all households are increased by $100 each, the average in-
crease in their food expenditures will be very close to $25.25.
Note that when b is positive, an increase in x will lead to an increase in y, and a decrease in
x will lead to a decrease in y. In other words, when b is positive, the movements in x and y are in
the same direction. Such a relationship between x and y is called a positive linear relationship.
The regression line in this case slopes upward from left to right. On the other hand, if the value
of b is negative, an increase in x will lead to a decrease in y, and a decrease in x will cause an
increase in y. The changes in x and y in this case are in opposite directions. Such a relationship
between x and y is called a negative linear relationship. The regression line in this case slopes
downward from left to right. The two diagrams in Figure 13.8 show these two cases.
For a regression model, b is computed as b = SS rv /SS„. The value of SS VV is always positive,
and that of SS rv can be positive or negative. Hence, the sign of b depends on the sign of SS XV .
If SS XV is positive (as in our example on the incomes and food expenditures of seven house-
holds), then b will be positive, and if SS„ is negative, then b will be negative.
Case Study 13-1 illustrates the difference between the population regression line and a sam-
ple regression line.
(a) Positive linear
relationship
(b) Negative linear
relationship
Figure 1 3.8 Positive and nega-
tive linear relationships between
x and y.
M Remember
REGRESSION
OF HEIGHTS
AND WEIGHTS
OF NBA
PLAYERS
Data Set III that accompanies this text lists the heights and weights of all National Basketball Association (NBA)
players who were on the rosters of all NBA teams as of May 2009. These data comprise the population of
NBA players for that point in time. We postulate the following simple linear regression model for these data:
y = A + Bx + e
where y is the weight (in pounds) and x is the height (in inches) of an NBA player.
Using the population data, we obtain the following regression line:
-281 + 6.32x
This equation gives the population regression line because it is obtained by using the population data.
(Note that in the population regression line we write ^ instead of /.) Thus, the true values of A and B
are, respectively,
A = -281 and B = 6.32
The value of 6 indicates that for every 1-inch increase in the height of an NBA player, weight in-
creases on average by 6.32 pounds. However, A = -281 does not make any sense. It states that the
Scatterplot and Regression Line of Weights and Heights for Population
350
300
S> 250
HI
3
200
150
•
•
*
• • • • ^
4
4
•
* * t i 1
, l i^*^ •
• 4
* ;
WW* * •
* • \
* *
• • ill! • • • •
>rP 1 « •
•
70
75
80
Height
85
90
Figure 13.9 Scatter diagram for the data on heights and weights of all NBA players.
13.2.4 Assumptions of the Regression Model
Like any other theory, the linear regression analysis is also based on certain assumptions. Con-
sider the population regression model
y = A + Bx + e (4)
Four assumptions are made about this model. These assumptions are explained next with refer-
ence to the example on incomes and food expenditures of households. Note that these assump-
tions are made about the population regression model and not about the sample regression model.
Assumption 1: The random error term e has a mean equal to zero for each x. In other words,
among all households with the same income, some spend more than the predicted food expen-
diture (and, hence, have positive errors) and others spend less than the predicted food expendi-
weight of a player with zero height is -281 pounds. (Recall from Section 13.2.3 that we must be very
careful if and when we apply the regression equation to predict y for values of x outside the range of
data used to find the regression line.) Figure 13.9 gives the scatter diagram and the regression line for
the heights and weights of all NBA players.
Next, we selected a random sample of 50 players and estimated the regression model for this sample.
The estimated regression line for the sample is
The values of a and b are
-267 + 6.1 3x
-267 and b = 6.13
These values of a and b give the estimates of A and 8 based on sample data. The scatter diagram and the re-
gression line for the sample observations on heights and weights is given in Figure 13.10. Note that this figure
does not show exactly 50 dots because some points/dots may be exactly the same or very close to each other.
Figure 13.10 Scatter diagram for the data on heights and weights of 50 NBA players.
As we can observe from Figures 13.9 and 13.10, the scatter diagrams for population and sample data
both show a (positive) linear relationship between the heights and weights of NBA players. Source: National Basketball Association.
ture (and, consequently, have negative errors). This assumption simply states that the sum of
the positive errors is equal to the sum of the negative errors, so that the mean of errors for all
households with the same income is zero. Thus, when the mean value of e is zero, the mean
value of y for a given x is equal to A + Bx, and it is written as
fjby\ x = A + Bx
As mentioned earlier in this chapter, fi y \ x is read as "the mean value of y for a given value of x."
When we find the values of A and B for model (4) using the population data, the points on the
regression line give the average values of y, denoted by /i ^ for the corresponding values of x.
Assumption 2: The errors associated with different observations are independent. According
to this assumption, the errors for any two households in our example are independent. In other
words, all households decide independently how much to spend on food.
575
576 Chapter 13 Simple Linear Regression
Assumption 3: For any given x, the distribution of errors is normal. The corollary of this as-
sumption is that the food expenditures for all households with the same income are normally
distributed.
Assumption 4: The distribution of population errors for each x has the same (constant) stan-
dard deviation, which is denoted by cr e . This assumption indicates that the spread of points
around the regression line is similar for all x values.
Figure 13.11 illustrates the meanings of the first, third, and fourth assumptions for house-
holds with incomes of $4000 and $7500 per month. The same assumptions hold true for any
other income level. In the population of all households, there will be many households with a
monthly income of $4000. Using the population regression line, if we calculate the errors for
all these households and prepare the distribution of these errors, it will look like the distribu-
tion given in Figure 13.11a. Its standard deviation will be a € . Similarly, Figure 13.11b gives the
distribution of errors for all those households in the population whose monthly income is $7500.
Its standard deviation is also er e . Both these distributions are identical. Note that the mean of
both of these distributions is E(e) = 0.
Normal distribution with (constant)
standard deviation a.
- Errors for households
with income = $4000
Normal distribution with (constant)
standard deviation a, —,
- Errors for households
with income = $7500
Figure 13.1 1 (a) Errors for households with an income of $4000 per month, (b) Errors for households
with an income of $7500 per month.
Figure 13.12 shows how the distributions given in Figure 13.11 look when they are plot-
ted on the same diagram with the population regression line. The points on the vertical line
through x = 40 give the food expenditures for various households in the population, each of
which has the same monthly income of $4000. The same is true about the vertical line through
x = 75 or any other vertical line for some other value of x.
13.2.5 A Note on the Use of Simple Linear Regression
We should apply linear regression with caution. When we use simple linear regression, we as-
sume that the relationship between two variables is described by a straight line. In the real
world, the relationship between variables may not be linear. Hence, before we use a simple
13.2 Simple Linear Regression Analysis 577
linear regression, it is better to construct a scatter diagram and look at the plot of the data
points. We should estimate a linear regression model only if the scatter diagram indicates such
a relationship. The scatter diagrams of Figure 13.13 give two examples for which the rela-
tionship between x and y is not linear. Consequently, fitting linear regression in such cases
would be wrong.
Figure 13.13 Nonlinear relationship
between x and y.
• • - •
• • •
.*
• • •
(/')
EXERCISES
CONCEPTS AND PROCEDURES
13.1 Explain the meaning of the words simple and linear as used in simple linear regression.
13.2 Explain the meaning of independent and dependent variables for a regression model.
13.3 Explain the difference between exact and nonexact relationships between two variables.
13.4 Explain the difference between linear and nonlinear relationships between two variables.
13.5 Explain the difference between a simple and a multiple regression model.
13.6 Briefly explain the difference between a deterministic and a probabilistic regression model.
13.7 Why is the random error term included in a regression model?
13.8 Explain the least squares method and least squares regression line. Why are they called by these
names?
13.9 Explain the meaning and concept of SSE. You may use a graph for illustration purposes.
13.10 Explain the difference between y and y,
13.11 Two variables x and y have a positive linear relationship. Explain what happens to the value of y
when x increases.
13.12 Two variables x and y have a negative linear relationship. Explain what happens to the value of y
when x increases.
13.13 Explain the following.
a. Population regression line
b. Sample regression line
c. True values of A and B
d. Estimated values of A and B that are denoted by a and b, respectively
13.14 Briefly explain the assumptions of the population regression model.
13.15 Plot the following straight lines. Give the values of the y-intercept and slope for each of these lines
and interpret them. Indicate whether each of the lines gives a positive or a negative relationship between
x and y.
a. y = 100 + 5x b. y = 400 - 4x
13.16 Plot the following straight lines. Give the values of the y-intercept and slope for each of these lines
and interpret them. Indicate whether each of the lines gives a positive or a negative relationship between
x and y.
a. y = -60 + 8x b. y = 300 - 6x
13.17 A population data set produced the following information.
N = 250, 2x = 9880, %y = 1456, %xy = 85,080, Xx 2 = 485,870
Find the population regression line.
578 Chapter 13 Simple Linear Regression
13.18 A population data set produced the following information.
N = 460, Xx = 3920, Xy = 2650, Xxy = 26,570, Xx 2 = 48,530
Find the population regression line.
13.19 The following information is obtained from a sample data set.
n = 10, Xx = 100, Xy = 220, Xxy = 3680, Xx 2 = 1140
Find the estimated regression line.
13.20 The following information is obtained from a sample data set.
n = 12, Xx = 66, Xy = 588, Xxy = 2244, Xx 2 = 396
Find the estimated regression line.
■ APPLICATIONS
13.21 A car rental company charges $50 a day and 20 cents per mile for renting a car. Let y be the total
rental charges (in dollars) for a car for one day and x be the miles driven. The equation for the relation-
ship between x and y is
y = 50 + .20x
a. How much will a person pay who rents a car for one day and drives it 100 miles?
b. Suppose each of 20 persons rents a car from this agency for one day and drives it 100 miles.
Will each of them pay the same amount for renting a car for a day or do you expect each per-
son to pay a different amount? Explain.
c. Is the relationship between x and y exact or nonexact?
13.22 Bob's Pest Removal Service specializes in removing wild creatures (skunks, bats, reptiles, etc.) from
private homes. He charges $70 to go to a house plus $20 per hour for his services. Let y be the total amount
(in dollars) paid by a household using Bob's services and x the number of hours Bob spends capturing
and removing the animal(s). The equation for the relationship between x and y is
y = 70 + 20x
a. Bob spent 3 hours removing a coyote from under Alice's house. How much will he be paid?
b. Suppose nine persons called Bob for assistance during a week. Strangely enough, each of these
jobs required exactly 3 hours. Will each of these clients pay Bob the same amount, or do you
expect each one to pay a different amount? Explain.
c. Is the relationship between x and y exact or nonexact?
13.23 A researcher took a sample of 25 electronics companies and found the following relationship be-
tween x and y, where x is the amount of money (in millions of dollars) spent on advertising by a company
in 2009 and y represents the total gross sales (in millions of dollars) of that company for 2009.
y = 3.6 + 11.75*
a. An electronics company spent $2 million on advertising in 2009. What are its expected gross
sales for 2009?
b. Suppose four electronics companies spent $2 million each on advertising in 2009. Do you ex-
pect these four companies to have the same actual gross sales for 2009? Explain.
c. Is the relationship between x and y exact or nonexact?
13.24 A researcher took a sample of 10 years and found the following relationship between x and y, where
x is the number of major natural calamities (such as tornadoes, hurricanes, earthquakes, floods, etc.) that
occurred during a year and y represents the average annual total profits (in millions of dollars) of a sample
of insurance companies in the United States.
y = 342.6 - 2.10.x
a. A randomly selected year had 24 major calamities. What are the expected average profits of U.S.
insurance companies for that year?
b. Suppose the number of major calamities was the same for each of 3 years. Do you expect
the average profits for all U.S. insurance companies to be the same for each of these 3 years?
Explain.
c. Is the relationship between x and y exact or nonexact?
13.25 An auto manufacturing company wanted to investigate how the price of one of its car models de-
preciates with age. The research department at the company took a sample of eight cars of this model and
collected the following information on the ages (in years) and prices (in hundreds of dollars) of these cars.
13.2 Simple Linear Regression Analysis 579
Age
8
3
6
9
2
5
6
3
Price
45
210
100
33
267
134
109
235
a. Construct a scatter diagram for these data. Does the scatter diagram exhibit a linear relationship
between ages and prices of cars?
b. Find the regression line with price as a dependent variable and age as an independent variable.
c. Give a brief interpretation of the values of a and b calculated in part b.
d. Plot the regression line on the scatter diagram of part a and show the errors by drawing verti-
cal lines between scatter points and the regression line.
e. Predict the price of a 7-year-old car of this model.
f. Estimate the price of an 18-year-old car of this model. Comment on this finding.
13.26 The following table gives information on the amount of sugar (in grams) and the calorie count in
one serving of a sample of 13 varieties of Kellogg's cereal.
Sugar (grams)
4
15
12
11 8
6
7
2
7
14
20
3
13
Calories
120
200
140
110 120
80
190
100
120
190
190
110
120
Source: kelloggs.com.
a. Construct a scatter diagram for these data. Does the scatter diagram exhibit a linear relationship
between the amount of sugar and the number of calories per serving?
b. Find the predictive regression equation of the number of calories on the amount of sugar.
c. Give a brief interpretation of the values of a and b calculated in part b.
d. Plot the predictive regression line on the scatter diagram of part a and show the errors by draw-
ing vertical lines between scatter points and the predictive regression line.
e. Calculate the predicted calorie count for a cereal with 16 grams of sugar per serving.
f. Estimate the calorie count for a cereal with 52 grams of sugar per serving. Comment on this finding.
13.27 A diabetic is interested in determining how the amount of aerobic exercise impacts his blood sugar.
When his blood sugar reaches 170 mg/dL, he goes out for a run at a pace of 10 minutes per mile. On dif-
ferent days, he runs different distances and measures his blood sugar after completing his run. Note: The
preferred blood sugar level is in the range of 80 to 120 mg/dL. Levels that are too low or too high are ex-
tremely dangerous. The data generated are given in the following table.
Distance (miles)
2
2
2.5
2.5
3
3
3.5
3.5
4
4
4.5
4.5
Blood sugar (mg/dL)
136
146
131
125
120
116
104
95
85
94
83
75
a. Construct a scatter diagram for these data. Does the scatter diagram exhibit a linear relationship
between distance run and blood sugar level?
b. Find the predictive regression equation of blood sugar level on the distance run.
c. Give a brief interpretation of the values of a and b calculated in part b.
d. Plot the predictive regression line on the scatter diagram of part a and show the errors by draw-
ing vertical lines between scatter points and the predictive regression line.
e. Calculate the predicted blood sugar level count after a run of 3.1 miles (5 kilometers).
f. Estimate the blood sugar level after a 10-mile run. Comment on this finding.
13.28 While browsing through the magazine rack at a bookstore, a statistician decides to examine the re-
lationship between the price of a magazine and the percentage of the magazine space that contains ad-
vertisements. The data are given in the following table.
Percentage containing ads
37
43
58
49
70
28
65
32
Price ($)
5.50
6.95
4.95
5.75
3.95
8.25
5.50
6.75
a. Construct a scatter diagram for these data. Does the scatter diagram exhibit a linear relationship
between the percentage of a magazine's space containing ads and the price of the magazine?
b. Find the estimated regression equation of price on the percentage containing ads.
c. Give a brief interpretation of the values of a and b calculated in part b.
d. Plot the estimated regression line on the scatter diagram of part a, and show the errors by draw-
ing vertical lines between scatter points and the predictive regression line.
e. Predict the price of a magazine with 50% of its space containing ads.
f. Estimate the price of a magazine with 99% of its space containing ads. Comment on this finding.
580 Chapter 13 Simple Linear Regression
13.29 The following table gives the total 2008 payroll (on the opening day of the season, rounded to the
nearest million dollars) and the number of runs scored during the 2008 season by each of the National
League baseball teams.
Total Payroll
Team
(millions of dollars)
Runs Scored
Arizona Diamondbacks
"7 1
Tin
/zU
Atlanta Braves
V /
/jj
Chicago Cubs
1 j J
OJJ
Cincinnati Reds
7 1
/U4
Colorado Rockies
~7C
IJ
7/1*7
Florida Marlins
11
J I
1 l\)
Houston Astros
1 ni
1 1Z
Los Angeles Dodgers
i on
1 uu
/uu
Milwaukee Brewers
80
750
New York Mets
136
799
Philadelphia Phillies
113
799
Pittsburgh Pirates
49
735
San Diego Padres
43
637
San Francisco Giants
82
640
St. Louis Cardinals
89
779
Washington Nationals
59
641
Source: ESPN.com.
a. Find the least squares regression line with total payroll as the independent variable and runs
scored as the dependent variable.
b. Is the equation of the regression line obtained in part a the population regression line? Why or why
not? Do the values of the ^'-intercept and the slope of the regression line give A and 5 or a and bl
c. Give a brief interpretation of the values of the y-intercept and the slope obtained in part a.
d. Predict the number of runs scored by a team with a total payroll of $84 million.
13.30 The following table gives the total 2008 payroll (on the opening day of the season, rounded to the
nearest million dollars) and the number of runs scored during the 2008 season by each of the American
League baseball teams.
Total Payroll
Team
(millions of dollars)
Runs Scored
Baltimore Orioles
67
782
Boston Red Sox
123
845
Chicago White Sox
96
811
Cleveland Indians
82
805
Detroit Tigers
115
821
Kansas City Royals
71
691
Los Angeles Angels
114
765
Minnesota Twins
65
829
New York Yankees
201
789
Oakland Athletics
62
646
Seattle Mariners
99
671
Tampa Bay Rays
63
774
Texas Rangers
69
901
Toronto Blue Jays
81
714
Source: ESPN.com.
13.3 Standard Deviation of Random Errors
a. Find the least squares regression line with total payroll as the independent variable and runs
scored as the dependent variable.
b. Is the equation of the regression line obtained in part a the population regression line? Why or
why not? Do the values of the y-intercept and the slope of the regression line give A and B or
a and £>?
c. Give a brief interpretation of the values of the y-intercept and the slope obtained in part a.
d. Predict the number of runs scored by a team with a total payroll of $84 million.
13.3 Standard Deviation of Random Errors
When we consider incomes and food expenditures, all households with the same income are
expected to spend different amounts on food. Consequently, the random error e will assume dif-
ferent values for these households. The standard deviation <j e measures the spread of these er-
rors around the population regression line. The standard deviation of errors tells us how widely
the errors and, hence, the values of y are spread for a given x. In Figure 13.12, which is repro-
duced as Figure 13.14, the points on the vertical line through x = 40 give the monthly food
expenditures for all households with a monthly income of $4000. The distance of each dot from
the point on the regression line gives the value of the corresponding error. The standard devia-
tion of errors <x e measures the spread of such points around the population regression line. The
same is true for x = 75 or any other value of x.
Note that a € denotes the standard deviation of errors for the population. However, usually
<j e is unknown. In such cases, it is estimated by s e , which is the standard deviation of errors for
the sample data. The following is the basic formula to calculate s e :
where SSE = X(y - yf
In this formula, n — 2 represents the degrees of freedom for the regression model. The reason
df = n — 2 is that we lose one degree of freedom to calculate x and one for y.
Degrees of Freedom for a Simple Linear Regression Model The degrees of freedom for a simple
linear regression model are
df = n — 2
For computational purposes, it is more convenient to use the following formula to calculate
the standard deviation of errors s e .
582
Chapter 13 Simple Linear Regression
Calculating the standard
deviation of errors.
Standard Deviation of Errors The standard deviation of errors is calculated as 3
where SS VV = %y
S S yy b S S
n - 2
(Sy) 2
n
The calculation of SS,. V was discussed earlier in this chapter.
4
Like the value of SS ct , the value of SS VV is always positive.
Example 13-2 illustrates the calculation of the standard deviation of errors for the data of
Table 13.1.
■ EXAMPLE 13-2
Compute the standard deviation of errors s e for the data on monthly incomes and food
expenditures of the seven households given in Table 13.1.
Solution To compute s e , we need to know the values of SS YV , SS™, and b. In Example 13-1,
we computed SS AV and b. These values are
SS tv = 447.5714 and b = .2525
To compute SS™, we calculate Xy 2 as shown in Table 13.3.
Table 13.3
Income
Food Expenditure
X
y
y 2
55
14
196
83
24
576
38
13
169
61
16
256
33
9
81
49
15
225
67
17
289
Xx = 386
Xy = 108
V = 1792
The value of SS VV is
, (Xy) 2 (108) 2
SS VV = Xy 2 - = 1792 - = 125.7143
n 7
Hence, the standard deviation of errors is
/SS W - FSS^ / 125.7143 - .2525(447.5714)
n-2 V 7-2
1.5939
13.4 Coefficient of Determination
We may ask the question: How good is the regression model? In other words: How well does
the independent variable explain the dependent variable in the regression model? The coefficient
of determination is one concept that answers this question.
3 If we have access to population data, the value of a f is calculated using the formula
SSyy - B SS„
N
4 The basic formula to calculate SS VV is £(>' — y) 2 .
1 3.4 Coefficient of Determination 583
For a moment, assume that we possess information only on the food expenditures of house-
holds and not on their incomes. Hence, in this case, we cannot use the regression line to pre-
dict the food expenditure for any household. As we did in earlier chapters, in the absence of a
regression model, we use y to estimate or predict every household's food expenditure. Conse-
quently, the error of prediction for each household is now given by y — y, which is the differ-
ence between the actual food expenditure of a household and the mean food expenditure. If we
calculate such errors for all households in the sample and then square and add them, the re-
sulting sum is called the total sum of squares and is denoted by SST. Actually SST is the same
as SS VV and is defined as
SST = SS VV = X(y - yf
However, for computational purposes, SST is calculated using the following formula.
Total Sum of Squares (SST) The total sum of squares, denoted by SST, is calculated as
SST - V - ^
n
Note that this is the same formula that we used to calculate SS VV .
The value of SS VV , which is 125.7143, was calculated in Example 13-2. Consequently, the value
of SST is
SST = 125.7143
From Example 13—1, y = 15.4286. Figure 13.15 shows the error for each of the seven house-
holds in our sample using the scatter diagram of Figure 13.4 and using y.
| 24
5 18
CL
8 12
T3
O
Li_ D
- t
1
Figure 13.15 Total errors.
\y= 15.4286
20
40 60 80 100
Income
Now suppose we use the simple linear regression model to predict the food expenditure of
each of the seven households in our sample. In this case, we predict each household's food ex-
penditure by using the regression line we estimated earlier in Example 13-1, which is
y = 1.5050 + .2525*
The predicted food expenditures, denoted by y, for the seven households are listed in Table 13.4.
Also given are the errors and error squares.
Table 13.4
X
y
y = 1.5050 + .2525*
e =y -y
e 2 = ( y - yf
55
14
15.3925
-1.3925
1.9391
83
24
22.4625
1.5375
2.3639
38
13
11.1000
1.9000
3.6100
61
16
16.9075
-.9075
.8236
33
9
9.8375
-.8375
.7014
49
15
13.8775
1.1225
1.2600
67
17
18.4225
-1.4225
2.0235
2e 2 = %(y - yf = 12.7215
584 Chapter 13 Simple Linear Regression
We calculate the values of y (given in the third column of Table 13.4) by substituting the
values of x in the estimated regression model. For example, the value of x for the first house-
hold is 55. Substituting this value of x in the regression equation, we obtain
y = 1.5050 + .2525(55) = 15.3925
Similarly we find the other values of y. The error sum of squares SSE is given by the sum of
the fifth column in Table 13.4. Thus,
SSE = S(y - yf = 12.7215
The errors of prediction for the regression model for the seven households are shown in
Figure 13.16.
Thus, from the foregoing calculations,
SST = 125.7143 and SSE = 12.7215
These values indicate that the sum of squared errors decreased from 125.7143 to 12.7215 when
we used y in place of y to predict food expenditures. This reduction in squared errors is called
the regression sum of squares and is denoted by SSR. Thus,
SSR = SST - SSE = 125.7143 - 12.7215 = 112.9928
The value of SSR can also be computed by using the formula
SSR = %(y - yf
Regression Sum of Squares (SSR) The regression sum of squares, denoted by SSR, is
SSR = SST - SSE
Thus, SSR is the portion of SST that is explained by the use of the regression model, and
SSE is the portion of SST that is not explained by the use of the regression model. The sum of
SSR and SSE is always equal to SST. Thus,
SST = SSR + SSE
The ratio of SSR to SST gives the coefficient of determination. The coefficient of deter-
mination calculated for population data is denoted by p 2 (p is the Greek letter rho), and the one
calculated for sample data is denoted by r 2 . The coefficient of determination gives the propor-
tion of SST that is explained by the use of the regression model. The value of the coefficient
of determination always lies in the range zero to one. The coefficient of determination can be
calculated by using the formula
, SSR SST - SSE
r = or
SST SST
However, for computational purposes, the following formula is more efficient to use to calcu-
late the coefficient of determination.
1 3.4 Coefficient of Determination 585
Coefficient of Determination The coefficient of determination, denoted by r 2 , represents the
proportion of SST that is explained by the use of the regression model. The computational for-
mula for r 2 is 5
&SS„.
r =
SS VV
and < r 2 < 1
Example 13-3 illustrates the calculation of the coefficient of determination for a sample
data set.
■ EXAMPLE 13-3
For the data of Table 13.1 on monthly incomes and food expenditures of seven households,
calculate the coefficient of determination.
Solution From earlier calculations made in Examples 13-1 and 13-2,
b = .2525, SS. tv = 447.5714, and SS VV = 125.7143
Hence,
&SS„ (.2525)(447.5714)
r z= 1 = i — L = .90
SS yy 125.7143
Thus, we can state that SST is reduced by approximately 90% (from 125.7143 to 12.7215)
when we use y, instead of y, to predict the food expenditures of households. Note that r 2 is
usually rounded to two decimal places. I
The total sum of squares SST is a measure of the total variation in food expenditures, the
regression sum of squares SSR is the portion of total variation explained by the regression model
(or by income), and the error sum of squares SSE is the portion of total variation not explained
by the regression model. Hence, for Example 13-3 we can state that 90% of the total variation
in food expenditures of households occurs because of the variation in their incomes, and the re-
maining 10% is due to randomness and other variables.
Usually, the higher the value of r 2 , the better is the regression model. This is so because if
r 2 is larger, a greater portion of the total errors is explained by the included independent vari-
able, and a smaller portion of errors is attributed to other variables and randomness.
EXERCISES
CONCEPTS AND PROCEDURES
13.31 What are the degrees of freedom for a simple linear regression model?
13.32 Explain the meaning of coefficient of determination.
13.33 Explain the meaning of SST and SSR. You may use graphs for illustration purposes.
13.34 A population data set produced the following information.
N = 250, 2* = 9880, 2y = 1456, Sxy = 85,080,
£x 2 = 485,870, and £>' 2 = 135,675
Find the values of cr f and p 2 .
5 If we have access to population data, the value of p 2 is calculated using the formula
, B ss v,
F SS„
The values of SS rv and SS vt used here are calculated for the population data set.
Calculating the
coefficient of determination.
586 Chapter 13 Simple Linear Regression
13.35 A population data set produced the following information.
N = 460, Xx = 3920, Xy = 2650, Sxy = 26,570,
Sx 2 = 48,530, and 2y 2 = 39,347
Find the values of a e and p 2 .
13.36 The following information is obtained from a sample data set.
n = 10, %x = 100, %y = 220, Sxy = 3680,
2x 2 = 1 140, and Xy 2 = 25,272
Find the values of s e and r.
13.37 The following information is obtained from a sample data set.
n = 12, %x = 66, 2y = 588, Xxy = 2244,
Xx 2 = 396, and 2y 2 = 58,734
Find the values of s e and r 2 .
■ APPLICATIONS
13.38 A container of one dozen large eggs was purchased at a local grocery store. Each egg was meas-
ured to determine its diameter (in millimeters) and weight (in grams). The results for the 12 eggs are given
in the following table.
Diameter (mm)
42.2
45.5
47.8
47.4
47.7
43.5
44.4
43.9
46.2
45.9
44.3
44.5
Weight (grams)
52.8
58.5
60.2
59.0
57.4
54.1
53.8
54.5
56.2
55.8
54.3
56.1
Find the following.
a. SS n ., SS™, and SS rv b. Standard deviation of errors
c. SST SSE, and SSR d. Coefficient of determination
13.39 The following table gives information on the number of megapixels and the prices of nine randomly
selected point-and-shoot digital cameras that were available on BestBuy.com on July 22, 2009.
Megapixels
10.3
10.2
7.0
9.1
10.0
12.1
8.0
5.0
14.7
Price ($)
130
150
62
160
200
280
125
60
400
Compute the following.
a. SS rv , SS, T , and SS VV b. Standard deviation of errors
c. SST, SSE, and SSR d. Coefficient of determination
13.40 Refer to Exercise 13.25. The following table, which gives the ages (in years) and prices (in hun-
dreds of dollars) of eight cars of a specific model, is reproduced from that exercise.
Age
8
3
6
9
2
5
6
3
Price
45
210
100
33
267
134
109
235
a. Calculate the standard deviation of errors.
b. Compute the coefficient of determination and give a brief interpretation of it.
13.41 The following table, reproduced from Exercise 13.26, gives information on the amount of sugar (in
grams) and the calorie count in one serving of a sample of 13 varieties of Kellogg's cereal.
Sugar (grams)
4
15
12
11
8 6 7
2
7
14
20
3
13
Calories
120
200
140
110
120 80 190
100
120
190
190
110
120
Source: kelloggs.com.
a. Determine the standard deviation of errors.
b. Find the coefficient of determination and give a brief interpretation of it.
13.5 Inferences About 8 587
13.42 The following table containing data on the aerobic exercise levels (running distance in miles) and
blood sugar levels for 12 different days for a diabetic is reproduced from Exercise 13.27.
Distance (miles)
2
2
2.5
2.5 3
3
3.5
3.5
4
4
4.5
4.5
Blood sugar (mg/dL)
136
146
131
125 120
116
104
95
85
94
83
75
a. Find the standard deviation of errors.
b. Compute the coefficient of determination. What percentage of the variation in blood sugar level
is explained by the least squares regression of blood sugar level on the distance run? What per-
centage of this variation is not explained?
13.43 The following table, reproduced from Exercise 13.28, lists the percentages of space for eight mag-
azines that contain advertisements and the prices of these magazines.
Percentage containing ads
37
43
58
49
70
28
65
32
Price ($)
5.50
6.95
4.95
5.75
3.95
8.25
5.50
6.75
a. Find the standard deviation of errors.
b. Compute the coefficient of determination. What percentage of the variation in price is explained
by the least squares regression of price on the percentage of magazine space containing ads?
What percentage of this variation is not explained?
13.44 Refer to data given in Exercise 13.29 on the total 2008 payroll and the number of runs scored dur-
ing the 2008 season by each of the National League baseball teams.
a. Find the standard deviation of errors, a e . (Note that this data set belongs to a population.)
b. Compute the coefficient of determination, p 2 .
13.45 Refer to data given in Exercise 13.30 on the total 2008 payroll and the number of runs scored dur-
ing the 2008 season by each of the American League baseball teams.
a. Find the standard deviation of errors, a e . (Note that this data set belongs to a population.)
b. Compute the coefficient of determination, p 2 .
13.5 Inferences About B
This section is concerned with estimation and tests of hypotheses about the population regression
slope B. We can also make confidence intervals and test hypotheses about the y-intercept A of the
population regression line. However, making inferences about A is beyond the scope of this text.
13.5.1 Sampling Distribution of b
One of the main purposes for determining a regression line is to find the true value of the slope
B of the population regression line. However, in almost all cases, the regression line is estimated
using sample data. Then, based on the sample regression line, inferences are made about the
population regression line. The slope b of a sample regression line is a point estimator of the
slope B of the population regression line. The different sample regression lines estimated for
different samples taken from the same population will give different values of b. If only one
sample is taken and the regression line for that sample is estimated, the value of b will depend
on which elements are included in the sample. Thus, b is a random variable, and it possesses a
probability distribution that is more commonly called its sampling distribution. The shape of
the sampling distribution of b, its mean, and standard deviation are given next.
Mean, Standard Deviation, and Sampling Distribution of b Because of the assumption of
normally distributed random errors, the sampling distribution of b is normal. The mean and stan-
dard deviation of b, denoted by jjb h and a h , respectively, are
fju b = B and a b =
588 Chapter 13 Simple Linear Regression
However, usually the standard deviation of population errors a e is not known. Hence, the
sample standard deviation of errors s e is used to estimate cr e . In such a case, when a € is un-
known, the standard deviation of b is estimated by s b , which is calculated as
If cr € is known, the normal distribution can be used to make inferences about B. However, if er E is
not known, the normal distribution is replaced by the t distribution to make inferences about B.
13.5.2 Estimation of B
The value of b obtained from the sample regression line is a point estimate of the slope B of
the population regression line. As mentioned in Section 13.5.1, if <x e is not known, the t distri-
bution is used to make a confidence interval for B.
Confidence Interval for B The (1 - a) 100% confidence interval for B is given by
b ± ts b
where s h
vss^
and the value of t is obtained from the t distribution table for a/2 area in the right tail of the t
distribution and n — 2 degrees of freedom.
Example 13-4 describes the procedure for making a confidence interval for B.
I EXAMPLE 13-4
Construct a 95% confidence interval for B for the data on incomes and food expenditures of
„. Constructing a seven households given in Table 13.1
confidence interval for a. °
Solution From the given information and earlier calculations in Examples 13-1 and 13-2,
n = 7, b = .2525, SS„ = 1772.8571, and s e = 1.5939
The confidence level is 95%. We have
s e 1.5939
s b = J— = , = .0379
VSS^ V1772.8571
df = n - 2 = 7- 2 = 5
a/2 = (1 - .95)/2 = .025
From the t distribution table, the value of t for 5 df and .025 area in the right tail of the t dis-
tribution curve is 2.571. The 95% confidence interval for B is
b ± ts b = .2525 ± 2.571(.0379) = .2525 ± .0974 = .155 to .350
Thus, we are 95% confident that the slope B of the population regression line is between
.155 and .350. ■
13.5.3 Hypothesis Testing About B
Testing a hypothesis about B when the null hypothesis is B = (that is, the slope of the re-
gression line is zero) is equivalent to testing that x does not determine y and that the regression
line is of no use in predicting y for a given x. However, we should remember that we are test-
ing for a linear relationship between x and y. It is possible that x may determine y nonlinearly.
Hence, a nonlinear relationship may exist between x and y.
To test the hypothesis that x does not determine y linearly, we will test the null hypothesis
that the slope of the regression line is zero; that is, B = 0. The alternative hypothesis can be:
(1) x determines y, that is, B # 0; (2) x determines y positively, that is, B > 0; or (3) x deter-
mines y negatively, that is, B < 0.
The procedure used to make a hypothesis test about B is similar to the one used in earlier
chapters. It involves the same five steps. Of course, use can use the /7-value approach too.
Test Statistic for b The value of the test statistic tfor b is calculated as
b — B
t =
The value of B is substituted from the null hypothesis.
Example 13-5 illustrates the procedure for testing a hypothesis about B.
M EXAMPLE 13-5
Test at the 1% significance level whether the slope of the regression line for the example on
incomes and food expenditures of seven households is positive.
Solution From the given information and earlier calculations in Examples 13-1 and 13 — 4,
n = 7, b = .2525, and s b = .0379
Step 1. State the null and alternative hypotheses.
We are to test whether or not the slope B of the population regression line is positive. Hence,
the two hypotheses are
H :B = (The slope is zero)
Hi'.B > (The slope is positive)
Note that we can also write the null hypothesis as H : B < 0, which states that the slope is
either zero or negative.
Step 2. Select the distribution to use.
Here, a € is not known. Hence, we will use the t distribution to make the test about B.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .01. The > sign in the alternative hypothesis indicates that the test
is right-tailed. Therefore,
Area in the right tail of the t distribution = a = .01
df=n - 2 = 1 - 2 = 5
From the t distribution table, the critical value of t for 5 df and .01 area in the right tail of the t
distribution is 3.365, as shown in Figure 13.17.
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for b is calculated as follows:
j- From H a
b-B .2525 -
t = = = 6.662
s b .0379
590 Chapter 13 Simple Linear Regression
Step 5. Make a decision.
The value of the test statistic t = 6.662 is greater than the critical value of t = 3.365, and
it falls in the rejection region. Hence, we reject the null hypothesis and conclude that x (in-
come) determines y (food expenditure) positively. That is, food expenditure increases with an
increase in income and it decreases with a decrease in income.
Using the p-Value to Make a Decision
We can find the range for the p- value (as we did in Chapters 9 and 10) from the t distribution
table, Table V of Appendix C, and make a decision by comparing that p-value with the sig-
nificance level. For this example, df = 5, and the observed value of t is 6.662. From Table V
(the t distribution table) in the row of df = 5, the largest value of t is 5.893 for which the area
in the right tail of the t distribution is .001. Since our observed value of t = 6.662 is larger
than 5.893, the p-value for t = 6.662 is less than .001, that is,
p-value < .001
Note that if we use technology to find this p-value, we will obtain a p-value of .000. Thus,
we can state that for any a equal to or higher than .001 (the upper limit of the p-value range),
we will reject the null hypothesis. For our example, a = .01, which is larger than the p-value
of .001. As a result, we reject the null hypothesis. I
Note that the null hypothesis does not always have to be B = 0. We may test the null hypoth-
esis that B is equal to a certain value. See Exercises 13.47 to 13.50, 13.54, 13.57, and 13.58 for
such cases.
1
EXERCISES
CONCEPTS AND PROCEDURES
13.46 Describe the mean, standard deviation, and shape of the sampling distribution of the slope b of the
simple linear regression model.
13.47 The following information is obtained for a sample of 16 observations taken from a population.
SS,.
340.700,
1.951, and
12.45 + 6.32*
Make a 99% confidence interval for B.
Using a significance level of .025, can you conclude that B is positive?
Using a significance level of .01, can you conclude that B is different from zero?
Using a significance level of .02, test whether B is different from 4.50. {Hint: The null hypothesis
here will be H a : B = 4.50, and the alternative hypothesis will be H { : B + 4.50. Notice that the
value of B = 4.50 will be used to calculate the value of the test statistic t.)
13.48 The following information is obtained for a sample of 25 observations taken from a population.
SS XV = 274.600, s e = .932, and y = 280.56 - 3.77*
a. Make a 95% confidence interval for B.
b. Using a significance level of .01, test whether B is negative.
c. Testing at the 5% significance level, can you conclude that B is different from zero?
d. Test if B is different from -5.20. Use a = .01.
13.5 Inferences About B 591
13.49 The following information is obtained for a sample of 100 observations taken from a population.
SS.„ = 524.884 s e = 1.464, and y = 5.48 + 2.50r
a. Make a 98% confidence interval for B.
b. Test at the 2.5% significance level whether B is positive.
c. Can you conclude that B is different from zero? Use a = .01.
d. Using a significance level of .01, test whether B is greater than 1.75.
13.50 The following information is obtained for a sample of 80 observations taken from a population.
SS„ = 380.592, s e = .961, and y = 160.24 - 2.70jc
a. Make a 97% confidence interval for B.
b. Test at the 1% significance level whether B is negative.
c. Can you conclude that B is different from zero? Use a = .01.
d. Using a significance level of .02, test whether B is less than —1.25.
■ APPLICATIONS
13.51 Refer to Exercise 13.25. The data on ages (in years) and prices (in hundreds of dollars) for eight
cars of a specific model are reproduced from that exercise.
Age
8
3
6
9
2
5
6
3
Price
45
210
100
33
267
134
109
235
a. Construct a 95% confidence interval for B. You can use results obtained in Exercises 13.25 and
13.40 here.
b. Test at the 5% significance level whether B is negative.
13.52 The data given in the table below are the midterm scores in a course for a sample of 10 students
and the scores of student evaluations of the instructor. (In the instructor evaluation scores, 1 is the lowest
and 4 is the highest score.)
Instructor score
3
2
3
1
2
4
3
4 4
2
Midterm score
90
75
97
64
47
99
75 8
8 93
81
a. Find the regression of instructor scores on midterm scores.
b. Construct a 99% confidence interval for B.
c. Test at the 1% significance level whether B is positive.
13.53 The following data give the experience (in years) and monthly salaries (in hundreds of dollars) of
nine randomly selected secretaries.
Experience
14
3
5
6
4
9
18
5
16
Monthly salary
62
29
37
43
35
60
67
32
60
a. Find the least squares regression line with experience as an independent variable and monthly
salary as a dependent variable.
b. Construct a 98% confidence interval for B.
c. Test at the 2.5% significance level whether B is greater than zero.
13.54 The following table, reproduced from Exercise 13.26, gives information on the amount of sugar (in
grams) and the calorie count in one serving of a sample of 13 varieties of Kellogg's cereal.
Sugar (grams)
4
15
12
11 8
6
7
2
7
14
20
3
13
Calories
120
200
140
110 120
80
190
100
120
190
190
110
120
Source: kelloggs.com.
a. Make a 95% confidence interval for B. You can use the calculations made in Exercises 13.26
and 13.41 here.
b. It is well known that each additional gram of carbohydrate adds 4 calories. Sugar is one type of
carbohydrate. Using regression equation for the data in the table, test at the 1% significance level
whether B is different from 4.
592 Chapter 13 Simple Linear Regression
13.55 The following table containing data on the aerobic exercise levels (running distance in miles) and
blood sugar levels for 12 different days for a diabetic is reproduced from Exercise 13.27.
Distance (miles)
2
2
2.5
2.5 3
3
3.5
3.5
4
4
4.5
4.5
Blood sugar (mg/dL)
136
146
131
125 120
116
104
95
85
94
83
75
a. Construct a 99% confidence interval for B. You can use the calculations made in Exercises 13.27
and 13.42 here.
b. Test at the 1% significance level whether B is negative.
13.56 The following table, reproduced from Exercise 13.28, lists the percentages of space for eight mag-
azines that contain advertisements and the prices of these magazines.
Percentage containing ads
37
43
58
49
70
28
65
32
Price ($)
5.50
6.95
4.95
5.75
3.95
8.25
5.50
6.75
a. Construct a 98% confidence interval for B. You can use the calculations made in Exercises 13.28
and 13.43 here.
b. Testing at the 5% significance level, can you conclude that B is different from zero?
13.57 Refer to Exercise 13.38. A container of one dozen large eggs was purchased at a local grocery store.
Each egg was measured to determine its diameter (in millimeters) and weight (in grams). The results for
the 12 eggs are given in the following table.
Diameter (mm)
42.2
45.5
47.8
47.4
47.7
43.5
44.4
43.9
46.2
45.9
44.3
44.5
Weight (grams)
52.8
58.5
60.2
59.0
57.4
54.1
53.8
54.5
56.2
55.8
54.3
56.1
a. Find the least squares regression equation with diameter as the independent variable and weight
as the dependent variable.
b. Make a 95% confidence interval for B. You may use the results obtained in Exercise 13.38 here.
c. A poultry specialist believes that B = 1.0 for the relationship between weight and diameter. Test
at the 5% significance level whether B is different from 1.0.
13.58 The following table, reproduced from Exercise 13.39, gives information on the number of megapixels
and the prices of nine randomly selected point-and-shoot digital cameras that were available on BestBuy.com
on July 22, 2009.
Megapixels
10.3
10.2
7.0
9.1
10.0
12.1
8.0
5.0
14.7
Price ($)
130
150
62
160
200
280
125
60
400
a. Find the regression equation y = a + bx, where x is the number of megapixels and y is the price.
You can use the results obtained in Exercise 13.39 here.
b. Construct a 98% confidence interval for B.
c. At the time when point-and-shoot digital cameras ranged between 3 and 8 megapixels, each ad-
ditional megapixel cost approximately $50. Test at the 2% significance level whether B is less
than $50.
13.6 Linear Correlation
This section describes the meaning and calculation of the linear correlation coefficient and the
procedure to conduct a test of hypothesis about it.
13.6.1 Linear Correlation Coefficient
Another measure of the relationship between two variables is the correlation coefficient. This
section describes the simple linear correlation, for short linear correlation, which measures
the strength of the linear association between two variables. In other words, the linear cor-
relation coefficient measures how closely the points in a scatter diagram are spread around
the regression line. The correlation coefficient calculated for the population data is denoted
13.6 Linear Correlation 593
by p (Greek letter rho) and the one calculated for sample data is denoted by r. (Note that
the square of the correlation coefficient is equal to the coefficient of determination.)
Value of the Correlation Coefficient The value of the correlation coefficient always lies in the
range — 1 to 1 ; that is,
- 1 < p < 1 and -1 < r < 1
Although we can explain the linear correlation using the population correlation coefficient p,
we will do so using the sample correlation coefficient r.
If r = 1, it is said to be a perfect positive linear correlation. In such a case, all points in
the scatter diagram lie on a straight line that slopes upward from left to right, as shown in Fig-
ure 13.18a. If r = —1, the correlation is said to be a perfect negative linear correlation. In this
case, all points in the scatter diagram fall on a straight line that slopes downward from left to
right, as shown in Figure 13.18£>. If the points are scattered all over the diagram, as shown in
Figure 13.18c, then there is no linear correlation between the two variables, and consequently
r is close to 0. Note that here r is not equal to zero but is very close to zero.
We do not usually encounter an example with perfect positive or perfect negative correlation.
What we observe in real-world problems is either a positive linear correlation with < r < 1 (that
is, the correlation coefficient is greater than zero but less than 1) or a negative linear correlation
with — 1 < r < (that is, the correlation coefficient is greater than — 1 but less than zero).
If the correlation between two variables is positive and close to 1, we say that the variables
have a strong positive linear correlation. If the correlation between two variables is positive but
close to zero, then the variables have a weak positive linear correlation. In contrast, if the cor-
relation between two variables is negative and close to — 1, then the variables are said to have a
strong negative linear correlation. If the correlation between two variables is negative but close
to zero, there exists a weak negative linear correlation between the variables. Graphically, a strong
correlation indicates that the points in the scatter diagram are very close to the regression line,
and a weak correlation indicates that the points in the scatter diagram are widely spread around
the regression line. These four cases are shown in Figure 13.19a-*/.
The linear correlation coefficient is calculated by using the following formula. (This cor-
relation coefficient is also called the Pearson product moment correlation coefficient.)
(«)
(c)
Figure 1 3.1 8 Linear correla-
tion between two variables.
(a) Perfect positive linear corre-
lation, r = 1 . (b) Perfect nega-
tive linear correlation, r = — 1.
(c) No linear correlation, r ~ 0.
Figure 1 3.1 9 Linear correlation
between two variables.
(c) Strong negative linear correlation (d) Weak negative linear correlation
(r is close to -1 ) (;- is negative and close to zero)
594 Chapter 13 Simple Linear Regression
Linear Correlation Coefficient The simple linear correlation coefficient, denoted by r, measures
the strength of the linear relationship between two variables for a sample and is calculated as 6
ss AT
\/SS vv SS VV
Because both SS vl . and SS VV are always positive, the sign of the correlation coefficient r
depends on the sign of SS tv . If SS AV is positive, then r will be positive, and if SS XV is negative,
then r will be negative. Another important observation to remember is that r and b, calculated
for the same sample, will always have the same sign. That is, both r and b are either positive
or negative. This is so because both r and b provide information about the relationship be-
tween x and y. Likewise, the corresponding population parameters p and B will always have
the same sign.
Example 13-6 illustrates the calculation of the linear correlation coefficient r.
Calculating the seV en households.
linear correlation coefficient.
EXAMPLE 13-6
Calculate the correlation coefficient for the example on incomes and food expenditures of
Solution From earlier calculations made in Examples 13-1 and 13-2,
SS, V = 447.5714, SS.„ = 1772.8571, and SS VV = 125.7143
Substituting these values in the formula for r, we obtain
ss « 447.5714
r = = = 95
VSS.„ SS VV V(1772.8571)(125.7143)
Thus, the linear correlation coefficient is .95. The correlation coefficient is usually rounded to
two decimal places. I
The linear correlation coefficient simply tells us how strongly the two variables are (lin-
early) related. The correlation coefficient of .95 for incomes and food expenditures of seven
households indicates that income and food expenditure are very strongly and positively corre-
lated. This correlation coefficient does not, however, provide us with any more information.
The square of the correlation coefficient gives the coefficient of determination, which
was explained in Section 13.4. Thus, (.95) 2 is .90, which is the value of r 2 calculated in
Example 13-3.
Sometimes the calculated value of r may indicate that the two variables are very strongly
linearly correlated, but in reality they are not. For example, if we calculate the correlation co-
efficient between the price of Coca-Cola and the size of families in the United States using data
for the last 30 years, we will find a strong negative linear correlation. Over time, the price of
Coca-Cola has increased and the size of families has decreased. This finding does not mean that
family size and the price of Coca-Cola are related. As a result, before we calculate the corre-
lation coefficient, we must seek help from a theory or from common sense to postulate whether
or not the two variables have a causal relationship.
Another point to note is that in a simple regression model, one of the two variables is cat-
egorized as an independent variable and the other is classified as a dependent variable. How-
ever, no such distinction is made between the two variables when the correlation coefficient is
calculated.
6 If we have access to population data, the value of p is calculated using the formula
SS tv
p ~ Vss„ ss rv
Here the values of SS,,„ SS [r , and SS VV are calculated using the population data.
13.6 Linear Correlation 595
13.6.2 Hypothesis Testing About the Linear
Correlation Coefficient
This section describes how to perform a test of hypothesis about the population correlation co-
efficient p using the sample correlation coefficient r. We can use the t distribution to make this
test. However, to use the t distribution, both variables should be normally distributed.
Usually (although not always), the null hypothesis is that the linear correlation coefficient
between the two variables is zero, that is, p = 0. The alternative hypothesis can be one of the
following: (1) the linear correlation coefficient between the two variables is less than zero, that
is, p < 0; (2) the linear correlation coefficient between the two variables is greater than zero,
that is, p > 0; or (3) the linear correlation coefficient between the two variables is not equal to
zero, that is, p # 0.
Test Statistic for r If both variables are normally distributed and the null hypothesis is
H : p = 0, then the value of the test statistic t is calculated as
In -2
Here n — 1 are the degrees of freedom.
Example 13-7 describes the procedure to perform a test of hypothesis about the linear cor-
relation coefficient.
■ EXAMPLE 13-7
Using the 1% level of significance and the data from Example 13—1, test whether the linear
correlation coefficient between incomes and food expenditures is positive. Assume that the
populations of both variables are normally distributed.
Solution From Examples 13-1 and 13-6,
n = 7 and r = .95
Below we use the five steps to perform this test of hypothesis.
Step 1. State the null and alternative hypotheses.
We are to test whether the linear correlation coefficient between incomes and food expen-
ditures is positive. Hence, the null and alternative hypotheses are, respectively,
H Q : p = (The linear correlation coefficient is zero.)
Hi. p > (The linear correlation coefficient is positive.)
Step 2. Select the distribution to use.
The population distributions for both variables are normally distributed. Hence, we can use
the t distribution to perform this test about the linear correlation coefficient.
Step 3. Determine the rejection and nonrejection regions.
The significance level is 1%. From the alternative hypothesis we know that the test is right-
tailed. Hence,
Area in the right tail of the t distribution = .01
df=n-2 = l - 2 = 5
From the t distribution table, the critical value of t is 3.365. The rejection and nonrejection re-
gions for this test are shown in Figure 13.20.
Performing a test of hypothesis
about the correlation coefficient.
596 Chapter 13 Simple Linear Regression
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for r is calculated as follows:
In - 2 I 7-2
t = rJ = .95 J — - = 6.803
V 1 - r 2 VI- (.95) 2
Step 5. Make a decision.
The value of the test statistic t = 6.803 is greater than the critical value of t = 3.365, and
it falls in the rejection region. Hence, we reject the null hypothesis and conclude that there is
a positive linear relationship between incomes and food expenditures.
Using the p-Value to Make a Decision
We can find the range for the /7-value from the t distribution table (Table V of Appendix C)
and make a decision by comparing that /?-value with the significance level. For this example,
df = 5, and the observed value of t is 6.803. From Table V (the t distribution table) in the row
of df = 5, the largest value of t is 5.893, for which the area in the right tail of the t distribu-
tion is .001. Since our observed value of t = 6.803 is larger than 5.893, the /j-value for t = 6.803
is less than .001, that is,
p-value < .001
Thus, we can state that for any a equal to or greater than .001 (the upper limit of the p-
value range), we will reject the null hypothesis. For our example, a = .01, which is greater
than the p-value of .001. As a result, we reject the null hypothesis. _
EXERCISES
CONCEPTS AND PROCEDURES
13.59 What does a linear correlation coefficient tell about the relationship between two variables? Within
what range can a correlation coefficient assume a value?
13.60 What is the difference between p and r? Explain.
13.61 Explain each of the following concepts. You may use graphs to illustrate each concept.
a. Perfect positive linear correlation
b. Perfect negative linear correlation
c. Strong positive linear correlation
d. Strong negative linear correlation
e. Weak positive linear correlation
f. Weak negative linear correlation
g. No linear correlation
13.62 Can the values of B and p calculated for the same population data have different signs? Explain.
13.63 For a sample data set, the linear correlation coefficient r has a positive value. Which of the fol-
lowing is true about the slope b of the regression line estimated for the same sample data?
a. The value of b will be positive.
b. The value of b will be negative.
c. The value of b can be positive or negative.
13.6 Linear Correlation 597
13.64 For a sample data set, the slope b of the regression line has a negative value. Which of the follow-
ing is true about the linear correlation coefficient r calculated for the same sample data?
a. The value of r will be positive.
b. The value of r will be negative.
c. The value of r can be positive or negative.
13.65 For a sample data set on two variables, the value of the linear correlation coefficient is zero. Does
this mean that these variables are not related? Explain.
13.66 Will you expect a positive, zero, or negative linear correlation between the two variables for each
of the following examples?
a. Grade of a student and hours spent studying
b. Incomes and entertainment expenditures of households
c. Ages of women and makeup expenses per month
d. Price of a computer and consumption of Coca-Cola
e. Price and consumption of wine
13.67 Will you expect a positive, zero, or negative linear correlation between the two variables for each
of the following examples?
a. SAT scores and CPAs of students
b. Stress level and blood pressure of individuals
c. Amount of fertilizer used and yield of corn per acre
d. Ages and prices of houses
e. Heights of husbands and incomes of their wives
13.68 A population data set produced the following information.
N = 250, tx = 9880, %y = 1456, 2xy = 85,080,
2x 2 = 485,870, and Xy 2 = 135,675
Find the linear correlation coefficient p.
13.69 A population data set produced the following information.
N = 460, Xx = 3920, %y = 2650, 2xy = 26,570,
Xx 2 = 48,530, and Xy 2 = 39,347
Find the linear correlation coefficient p.
13.70 A sample data set produced the following information.
n = 10, 2x = 100, Xy = 220, Xxy = 3680,
Xx 2 = 1 140, and Xy 2 = 25,272
a. Calculate the linear correlation coefficient r.
b. Using the 2% significance level, can you conclude that p is different from zero?
13.71 A sample data set produced the following information.
ii = 12, Xx = 66, Xy = 588, Xxy = 2244,
Xx 2 = 396, and Xy 2 = 58,734
a. Calculate the linear correlation coefficient r.
b. Using the 1% significance level, can you conclude that p is negative?
■ APPLICATIONS
13.72 Refer to Exercise 13.25. The data on ages (in years) and prices (in hundreds of dollars) for eight
cars of a specific model are reproduced from that exercise.
Age
8
3
6
9
2
5
6
3
Price
45
210
100
33
267
134
109
235
a. Do you expect the ages and prices of cars to be positively or negatively related? Explain.
b. Calculate the linear correlation coefficient.
e. Test at the 2.5% significance level whether p is negative.
598 Chapter 13 Simple Linear Regression
13.73 The following table, reproduced from Exercise 13.53, gives the experience (in years) and monthly
salaries (in hundreds of dollars) of nine randomly selected secretaries.
Experience
14
3
5
6
4
9
18
5
16
Monthly salary
62
29
37
43
35
60
67
32
60
a. Do you expect the experience and monthly salaries to be positively or negatively related? Explain.
b. Compute the linear correlation coefficient.
c. Test at the 5% significance level whether p is positive.
13.74 The following table lists the midterm and final exam scores for seven students in a statistics class.
Midterm score
79
95
81
66
87
94
59
Final exam score
85
97
78
76
94
84
67
a. Do you expect the midterm and final exam scores to be positively or negatively related?
b. Plot a scatter diagram. By looking at the scatter diagram, do you expect the correlation coefficient
between these two variables to be close to zero, 1, or —1?
c. Find the correlation coefficient. Is the value of r consistent with what you expected in parts a
and b?
d. Using the 1% significance level, test whether the linear correlation coefficient is positive.
13.75 The following data give the ages (in years) of husbands and wives for six couples.
Husband's age
43
57
28
19
35
39
Wife's age
37
51
32
20
33
38
a. Do you expect the ages of husbands and wives to be positively or negatively related?
b. Plot a scatter diagram. By looking at the scatter diagram, do you expect the correlation coeffi-
cient between these two variables to be close to zero, 1, or —1?
c. Find the correlation coefficient. Is the value of r consistent with what you expected in parts a
and b?
d. Using the 5% significance level, test whether the correlation coefficient is different from zero.
13.76 The following table, reproduced from Exercise 13.26, gives information on the amount of sugar (in
grams) and the calorie count in one serving of a sample of 13 varieties of Kellogg's cereal.
Sugar (grams)
4
15
12
11 8
6
7
2
7
14
20
3
13
Calories
120
200
140
110 120
80
190
100
120
190
190
110
120
Source: kelloggs.com.
a. Find the correlation coefficient. Is its sign the same as that of b calculated in Exercise 13.26?
b. Test at the 1% significance level whether the linear correlation coefficient between the two vari-
ables listed in the table is positive.
13.77 The following table, reproduced from Exercise 13.39, gives information on the number of
megapixels and the prices of nine randomly selected point-and-shoot digital cameras that were available
on BestBuy.com on July 22, 2009.
Megapixels
10.3
10.2
7.0
9.1
10.0
12.1
8.0
5.0
14.7
Price ($)
130
150
62
160
200
280
125
60
400
a. Find the correlation coefficient. Is the sign of the correlation coefficient the same as that of b
calculated in Exercise 13.58?
b. Test at the 1% significance level whether p is different from zero.
13.78 Refer to data given in Exercise 13.29 on the total 2008 payroll and the runs scored during
the 2008 season by each of the National League baseball teams. Compute the linear correlation coef-
ficient, p. Does it make sense to make a confidence interval and to test a hypothesis about p here? Explain.
13.79 Refer to data given in Exercise 13.30 on the total 2008 payroll and the runs scored during the 2008
season by each of the American League baseball teams. Compute the linear correlation coefficient, p. Does
it make sense to make a confidence interval and to test a hypothesis about p here? Explain.
13.7 Regression Analysis: A Complete Example 599
13.7 Regression Analysis: A Complete Example
This section works out an example that includes all the topics we have discussed so far in this
chapter.
■ EXAMPLE 13-8
A random sample of eight drivers insured with a company and having similar minimum re-
quired auto insurance policies was selected. The following table lists their driving experiences A complete example
(in years) and monthly auto insurance premiums (in dollars) of regression analysis.
Driving Experience
(years)
Monthly Auto Insurance
Premium ($)
5
64
2
87
12
50
9
71
15
44
6
56
25
42
16
60
(a) Does the insurance premium depend on the driving experience, or does the driving
experience depend on the insurance premium? Do you expect a positive or a negative
relationship between these two variables?
(b) Compute SS,,., SS VV , and SS VV .
(c) Find the least squares regression line by choosing appropriate dependent and inde-
pendent variables based on your answer in part a.
(d) Interpret the meaning of the values of a and b calculated in part c.
(e) Plot the scatter diagram and the regression line.
(f) Calculate r and r 2 , and explain what they mean.
(g) Predict the monthly auto insurance premium for a driver with 10 years of driving
experience.
(h) Compute the standard deviation of errors.
(i) Construct a 90% confidence interval for B.
(j) Test at the 5% significance level whether B is negative,
(k) Using a = .05, test whether p is different from zero.
Solution
(a) Based on theory and intuition, we expect the insurance premium to depend on driv-
ing experience. Consequently, the insurance premium is a dependent variable and
driving experience is an independent variable in the regression model. A new driver
is considered a high risk by the insurance companies, and he or she has to pay a
higher premium for auto insurance. On average, the insurance premium is expected
to decrease with an increase in the years of driving experience. Therefore, we ex-
pect a negative relationship between these two variables. In other words, both the
population correlation coefficient p and the population regression slope B are ex-
pected to be negative.
600 Chapter 13 Simple Linear Regression
(b) Table 13.5 shows the calculation of Xx, Xy, Xxy, Xx 2 , and Xy 2 .
Table 13.5
Experience
Premium
X
y
xy
x 1
y 1
j
ftA
2
87
174
4
7569
12
50
600
144
2500
9
71
639
81
5041
15
44
660
225
1936
6
56
336
36
3136
25
42
1050
625
1764
16
60
960
256
3600
2.x = 90
%y = 474
%xy = 4739
2x 2 = 1396
V = 29,642
(d)
(e)
The values of x and y are
x = Xx/n = 90/8 = 11.25
y = %y/n = 474/8 = 59.25
The values of SS™ SS„, and SS vr are computed as follows:
(Xx)(Xy) (90)(474)
SS„ = Xxy - = 4739 - = -593.5000
SS„ = Xx 2 -
(Xx)
1396
(90) 2
383.5000
, (Xy) 2 (474) 2
SS VV = Xy 2 - = 29,642 - = 1557.5000
n 8
(c) To find the regression line, we calculate a and b as follows:
SS t
SS,.
-593.5000
383.5000
-1.5476
a=y - bx = 59.25 - (- 1.5476)(1 1.25) = 76.6605
Thus, our estimated regression line y = a + bx is
y = 76.6605 - 1.5476x
The value of a = 76.6605 gives the value of y for x = 0; that is, it gives the monthly
auto insurance premium for a driver with no driving experience. However, as men-
tioned earlier in this chapter, we should not attach much importance to this statement
because the sample contains drivers with only 2 or more years of experience. The
value of b gives the change in y due to a change of one unit in x. Thus, b = — 1.5476
indicates that, on average, for every extra year of driving experience, the monthly
auto insurance premium decreases by $1.55. Note that when b is negative, y decreases
as x increases.
Figure 13.21 shows the scatter diagram and the regression line for the data on eight
auto drivers. Note that the regression line slopes downward from left to right. This re-
sult is consistent with the negative relationship we anticipated between driving expe-
rience and insurance premium.
13.7 Regression Analysis: A Complete Example 601
(f) The values of r and r 2 are computed as follows:
SS. VV -593.5000
VSS.„ SS VV V(383.5000)(1557.5000)
bSS xy (-1.5476)(-593.5000)
-.77
SS VV 1557.5000
.59
The value of r = — .77 indicates that the driving experience and the monthly auto in-
surance premium are negatively related. The (linear) relationship is strong but not very
strong. The value of r 2 = .59 states that 59% of the total variation in insurance
premiums is explained by years of driving experience, and 41% is not. The low value
of r 2 indicates that there may be many other important variables that contribute to
the determination of auto insurance premiums. For example, the premium is expected
to depend on the driving record of a driver and the type and age of the car.
(g) Using the estimated regression line, we find the predicted value of y for x = 10 is
y = 76.6605 - 1.5476* = 76.6605 - 1.5476(10) = $61.18
Thus, we expect the monthly auto insurance premium of a driver with 10 years of
driving experience to be $61.18.
(h) The standard deviation of errors is
SS„ - b SS tv 1557.5000 - (-1.5476)( -593.5000)
v - - = 10.3199
8-2
(i) To construct a 90% confidence interval for B, first we calculate the standard devia-
tion of b:
s e 10.3199
s b = ' = . = .5270
VSS^ V383.5000
For a 90% confidence level, the area in each tail of the t distribution is
a/2 = (1 - .90)/2 = .05
The degrees of freedom are
df=n - 2 = 8- 2 = 6
From the t distribution table, the t value for .05 area in the right tail of the t distribu-
tion and 6 dfis 1.943. The 90% confidence interval for B is
b ± ts b = -1.5476 ± 1.943(.5270)
= -1.5476 ± 1.0240 = -2.57 to -.52
Thus, we can state with 90% confidence that B lies in the interval —2.57 to —.52.
That is, on average, the monthly auto insurance premium of a driver decreases by an
amount between $.52 and $2.57 for every extra year of driving experience.
602 Chapter 13 Simple Linear Regression
(j) We perform the following five steps to test the hypothesis about B.
Step 1. State the null and alternative hypotheses.
The null and alternative hypotheses are, respectively,
H : B = (B is not negative.)
H { : B < {B is negative.)
Note that the null hypothesis can also be written as H : B > 0.
Step 2. Select the distribution to use.
Because cr e is not known, we use the t distribution to make the hypothesis test.
Step 3. Determine the rejection and nonrejection regions.
The significance level is .05. The < sign in the alternative hypothesis indicates that
it is a left-tailed test.
Area in the left tail of the t distribution = a = .05
df = n - 2 = 8- 2 = 6
From the t distribution table, the critical value of t for .05 area in the left tail of the t
distribution and 6 df is —1.943, as shown in Figure 13.22.
Figure 13.22
Reject H — *~
-« — Do not reject H
a = .05 -i
-1.943 t
t
Critical value of t
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for b is calculated as follows:
b — B
t =
Step 5. Make a decision.
The value of the test statistic t = —2.937 falls in the rejection region. Hence, we re-
ject the null hypothesis and conclude that B is negative. That is, the monthly auto in-
surance premium decreases with an increase in years of driving experience.
Using the p-Value to Make a Decision
We can find the range for the /7-value from the t distribution table (Table V of Ap-
pendix C) and make a decision by comparing that p-value with the significance level.
For this example, df = 6 and the observed value of t is —2.937. From Table V (the t
distribution table) in the row of df = 6, 2.937 is between 2.447 and 3.143. The corre-
sponding areas in the right tail of the t distribution are .025 and .01, respectively. Our
test is left-tailed, however, and the observed value of t is negative. Thus, t = —2.937
lies between —2.447 and —3.143. The corresponding areas in the left tail of the t dis-
tribution are .025 and .01. Therefore the range of the p-value is
.01 < ;>value < .025
Thus, we can state that for any a equal to or greater than .025 (the upper limit of the
p-value range), we will reject the null hypothesis. For our example, a = .05, which is
From H (
1.5476 -
= -2.937
.5270
13.7 Regression Analysis: A Complete Example 603
greater than the upper limit of the p-value of .025. As a result, we reject the null hy-
pothesis.
Note that if we use technology to find this p-value, we will obtain a p-value of .013.
Then we can reject the null hypothesis for any a > .013.
(k) We perform the following five steps to test the hypothesis about the linear correlation
coefficient p.
Step 1. State the null and alternative hypotheses.
The null and alternative hypotheses are, respectively,
H : p = (The linear correlation coefficient is zero.)
H x : p # (The linear correlation coefficient is different from zero.)
Step 2. Select the distribution to use.
Assuming that variables x and y are normally distributed, we will use the t dis-
tribution to perform this test about the linear correlation coefficient.
Step 3. Determine the rejection and nonrejection regions.
The significance level is 5%. From the alternative hypothesis we know that the test is
two-tailed. Hence,
Area in each tail of the t distribution
df=n - 2 = 8- 2 = 6
.05/2 = .025
From the t distribution table, Table V of Appendix C, the critical values of t are —2.447
and 2.447. The rejection and nonrejection regions for this test are shown in Figure 13.23.
Reject H
Do not reject H
Two critical values of t
Figure 13.23
Reject H
.025
Step 4. Calculate the value of the test statistic.
The value of the test statistic t for r is calculated as follows:
.77)
1 - (-.77) 2
-2.956
Step 5. Make a decision.
The value of the test statistic t = —2.956 falls in the rejection region. Hence, we re-
ject the null hypothesis and conclude that the linear correlation coefficient between
driving experience and auto insurance premium is different from zero.
Using the p-Value to Make a Decision
We can find the range for the /?-value from the t distribution table and make a decision by
comparing that /7-value with the significance level. For this example, df = 6 and the ob-
served value of t is —2.956. From Table V (the t distribution table) in the row of df = 6,
t = 2.956 is between 2.447 and 3.143. The corresponding areas in the right tail of the t
distribution curve are .025 and .01, respectively. Since the test is two tailed, the range of
the p-value is
2(.01) < p-value < 2(.025) or .02 < />value < .05
604 Chapter 13 Simple Linear Regression
Thus, we can state that for any a equal to or greater than .05 (the upper limit of the
p-value range), we will reject the null hypothesis. For our example, a = .05, which is
equal to the upper limit of the p-value. As a result, we reject the null hypothesis. ■
EXERCISES
APPLICATIONS
13.80 The owner of a small factory that produces working gloves is concerned about the high cost of air
conditioning in the summer but is afraid that keeping the temperature in the factory too high will lower
productivity. During the summer, he experiments with temperature settings from 68°F to 81°F and meas-
ures each day's productivity. The following table gives the temperature and the number of pairs of gloves
(in hundreds) produced on each of the 8 randomly selected days.
Temperature (°F)
72
71
78
75
81
77
68
76
Pairs of gloves
37
37
32
36
33
35
39
34
a. Do the pairs of gloves produced depend on temperature, or does temperature depend on pairs of
gloves produced? Do you expect a positive or a negative relationship between these two variables?
b. Taking temperature as an independent variable and pairs of gloves produced as a dependent vari-
able, compute SS„, SS,. V , and SS rv .
c. Find the least squares regression line.
d. Interpret the meaning of the values of a and b calculated in part c.
e. Plot the scatter diagram and the regression line.
f. Calculate r and r 2 , and explain what they mean.
g. Compute the standard deviation of errors.
h. Predict the number of pairs of gloves produced when x = 74.
i. Construct a 99% confidence interval for B.
j. Test at the 5% significance level whether B is negative,
k. Using a = .01 can you conclude that p is negative?
13.81 The following table gives information on the limited tread warranties (in thousands of miles) and
the prices of 12 randomly selected tires at a national tire retailer as of July 2009.
Warranty (thousands of miles)
60 70 75 50 80 55
65 65
70 65
60
65
Price per tire ($)
95 70 94 90 121 70
84 80
92 79
66
95
a. Taking warranty length as an independent variable and price per tire as a dependent variable,
compute SS vr , SS VV , and SS 1T .
b. Find the regression of price per tire on warranty length.
c. Briefly explain the meaning of the values of a and b calculated in part b.
d. Calculate r and r and explain what they mean.
e. Plot the scatter diagram and the regression line.
f. Predict the price of a tire with a warranty length of 73,000 miles.
g. Compute the standard deviation of errors.
h. Construct a 95% confidence interval for B.
i. Test at the 5% significance level if B is positive.
j. Using a = .025, can you conclude that the linear correlation coefficient is positive?
13.82 The recommended air pressure in a basketball is between 7 and 9 pounds per square inch (psi).
When dropped from a height of 6 feet, a properly inflated basketball should bounce upward between 52
and 56 inches (http://www.bestsoccerbuys.com/balls-basketball.html). The basketball coach at a local high
school purchased 10 new basketballs for the upcoming season, inflated the balls to pressures between
7 and 9 psi, and performed the bounce test mentioned above. The data obtained are given in the fol-
lowing table.
Pressure (psi)
7.8 8.1 8.3
7.4
8.9
7.2
8.6
7.5
8.1
8.5
Bounce height (inches)
54.1 54.3 55.2
53.3
55.4
52.2
55.7
54.6
54.8
55.3
13.7 Regression Analysis: A Complete Example 605
a. With the pressure as an independent variable and bounce height as a dependent variable, com-
pute SS,,, SS VV , and SS. n ,
b. Find the least squares regression line.
c. Interpret the meaning of the values of a and b calculated in part b.
d. Calculate r and r and explain what they mean.
e. Compute the standard deviation of errors.
f. Predict the bounce height of a basketball for x = 8.0.
g. Construct a 98% confidence interval for B.
h. Test at the 5% significance level whether B is different from zero.
i. Using a =.05, can you conclude that p is different from zero?
13.83 The following table gives information on the incomes (in thousands of dollars) and charitable con-
tributions (in hundreds of dollars) for the last year for a random sample of 10 households.
Income Charitable Contributions
76 15
57 4
140 42
97 33
75 5
107 32
65 10
77 18
102 28
53 4
a. With income as an independent variable and charitable contributions as a dependent variable,
compute SS„, SS, T , and SS„.
b. Find the regression of charitable contributions on income.
c. Briefly explain the meaning of the values of a and b.
d. Calculate r and r 2 and briefly explain what they mean.
e. Compute the standard deviation of errors.
f. Construct a 99% confidence interval for B.
g. Test at the 1 % significance level whether B is positive.
h. Using the 1% significance level, can you conclude that the linear correlation coefficient is dif-
ferent from zero?
13.84 The following data give information on the lowest cost ticket price (in dollars) and the average at-
tendance (rounded to the nearest thousand) for the last year for six football teams.
Ticket price
38.50
26.50
34.00
45.50
59.50
36.00
Attendance
56
65
71
69
55
42
a. Taking ticket price as an independent variable and attendance as a dependent variable, compute
SS„, SS VV , and SS„.
b. Find the least squares regression line.
c. Briefly explain the meaning of the values of a and b calculated in part b.
d. Calculate r and r 2 and briefly explain what they mean.
e. Compute the standard deviation of errors.
f. Construct a 90% confidence interval for B.
g. Test at the 2.5% significance level whether B is negative.
h. Using the 2.5% significance level, test whether p is negative.
13.85 The following table gives information on GPAs and starting salaries (rounded to the nearest thou-
sand dollars) of seven recent college graduates.
CPA
2.90
3.81
3.20
2.42
3.94
2.05
2.25
Starting salary
48
53
50
37
65
32
37
606
Chapter 13 Simple Linear Regression
a. With GPA as an independent variable and starting salary as a dependent variable, compute SS,
SS W , and SS_ VV .
b. Find the least squares regression line.
c. Interpret the meaning of the values of a and b calculated in part b.
d. Calculate r and r 2 and briefly explain what they mean.
e. Compute the standard deviation of errors.
f. Construct a 95% confidence interval for B.
g. Test at the 1% significance level whether B is different from zero.
h. Test at the 1% significance level whether p is positive.
13.8 Using the Regression Model
Let us return to the example on incomes and food expenditures to discuss two major uses of a
regression model:
1. Estimating the mean value of y for a given value of x. For instance, we can use our food
expenditure regression model to estimate the mean food expenditure of all households with
a specific income (say, $5500 per month).
2. Predicting a particular value of y for a given value of x. For instance, we can determine the
expected food expenditure of a randomly selected household with a particular monthly in-
come (say, $5500) using our food expenditure regression model.
13.8.1 Using the Regression Model for Estimating
the Mean Value of y
Our population regression model is
y = A + Bx + e
As mentioned earlier in this chapter, the mean value of y for a given x is denoted by [x y | x , read
as "the mean value of y for a given value of x." Because of the assumption that the mean value
of e is zero, the mean value of y is given by
fi y i x = A + Bx
Our objective is to estimate this mean value. The value of y, obtained from the sample regres-
sion line by substituting the value of x, is the point estimate of ix y | x for that x.
For our example on incomes and food expenditures, the estimated sample regression line
(from Example 13-1) is
y = 1.5050 + .2525x
Suppose we want to estimate the mean food expenditure for all households with a monthly in-
come of $5500. We will denote this population mean by /x^ r=55 or /a v | 55 . Note that we have
written x = 55 and not x = 5500 in /j. v | 55 because the units of measurement for the data used
to estimate the above regression line in Example 13-1 were hundreds of dollars. Using the re-
gression line, we find that the point estimate of fi y \ 55 is
y = 1.5050 + .2525(55) = $15.3925 hundred
Thus, based on the sample regression line, the point estimate for the mean food expenditure
fi y \ 55 for all households with a monthly income of $5500 is $1539.25 per month.
However, suppose we take a second sample of seven households from the same population
and estimate the regression line for this sample. The point estimate of fi y 1 55 obtained from the
regression line for the second sample is expected to be different. All possible samples of the
same size taken from the same population will give different regression lines as shown in Fig-
ure 13.24, and, consequently, a different point estimate of yu, v c . Therefore, a confidence inter-
val constructed for fi y | , based on one sample will give a more reliable estimate of fi y | v than
will a point estimate.
13.8 Using the Regression Model 607
To construct a confidence interval for fiy\„ we must know the mean, the standard devia-
tion, and the shape of the sampling distribution of its point estimator y.
The point estimator y of fly | v is normally distributed with a mean of A + Bx and a stan-
dard deviation of
[\ fa ~ xf
o"s = cr. a / — I
V n SS
where cr« is the standard deviation of y when it is used to estimate j± y | c , x is the value of x for
which we are estimating j± y | and cr € is the population standard deviation of e.
However, usually cr e is not known. Rather, it is estimated by the standard deviation of
sample errors s e . In this case, we replace cr € by s e and <x- by in the foregoing expression.
To make a confidence interval for /u.^ ,, we use the t distribution because cr f is not known.
Confidence Interval for fJL y \ x The (1 — a)100% confidence interval for fji y \ x for x = x is
y ± ts y,„
where the value of t is obtained from the t distribution table for a/2 area in the right tail of the
f distribution curve and df = n — 2. The value of is calculated as follows:
1 fa - W
n SS„
Example 13-9 illustrates how to make a confidence interval for the mean value of y, (x y \ x .
■ EXAMPLE 13-9
Refer to Example 13-1 on incomes and food expenditures. Find a 99% confidence interval for
the mean food expenditure for all households with a monthly income of $5500.
Solution Using the regression line estimated in Example 13-1, we find the point estimate
of the mean food expenditure for x = 55 as
y = 1.5050 + .2525(55) = $15.3925 hundred
The confidence level is 99%. Hence, the area in each tail of the t distribution is
a/2 = (1 -.99)/2 = .005
The degrees of freedom are
d/=n-2 = 7- 2 = 5
From the t distribution table, the t value for .005 area in the right tail of the t distribution and
5 df is 4.032. From calculations in Examples 13-1 and 13-2, we know that
Constructing a
confidence interval for the
mean value of y.
1.5939,
55.1429, and SS,.
1772.8571
608 Chapter 13 Simple Linear Regression
The standard deviation of y as an estimate of fi y \ x for x = 55 is calculated as follows:
1 (x -xf /l (55 - 55.1429)
\2
+ = (1.5939) A - + : = .6025
n SS,, V ; V 7 1772.8571
Hence, the 99% confidence interval for /j, y \ 55 is
y ± ts ym = 15.3925 ± 4.032(.6025)
= 15.3925 ± 2.4293 = 12.9632 to 17.8218
Thus, with 99% confidence we can state that the mean food expenditure for all households
with a monthly income of $5500 is between $1296.32 and $1782.18. ■
13.8.2 Using the Regression Model for Predicting
a Particular Value of y
The second major use of a regression model is to predict a particular value of y for a given value
of x — say, x . For example, we may want to predict the food expenditure of a randomly selected
household with a monthly income of $5500. In this case, we are not interested in the mean food ex-
penditure of all households with a monthly income of $5500 but in the food expenditure of one par-
ticular household with a monthly income of $5500. This predicted value of y is denoted by y p . Again,
to predict a single value of y for x = x from the estimated sample regression line, we use the value
of y as a point estimate ofy p . Using the estimated regression line, we find that y for x = 55 is
y = 1.5050 + .2525(55) = $15.3925 hundred
Thus, based on our regression line, the point estimate for the food expenditure of a given house-
hold with a monthly income of $5500 is $1539.25 per month. Note that y = 1539.25 is the point
estimate for the mean food expenditure for all households with x = 55 as well as for the pre-
dicted value of food expenditure of one household with x = 55.
Different regression lines estimated by using different samples of seven households each
taken from the same population will give different values of the point estimator for the pre-
dicted value of y for x = 55. Hence, a confidence interval constructed for y p based on one sam-
ple will give a more reliable estimate of y p than will a point estimate. The confidence interval
constructed for y p is more commonly called a prediction interval.
The procedure for constructing a prediction interval for y p is similar to that for construct-
ing a confidence interval for p, y | x except that the standard deviation of y is larger when we pre-
dict a single value of y than when we estimate fi y | ,.
The point estimator y of y p is normally distributed with a mean of A + Bx and a standard
deviation of
I + I + (*° ~ i)2
n SS„
where er ; is the standard deviation of the predicted value of y, x is the value of x for which we
are predicting y, and a € is the population standard deviation of e.
However, usually cr e is not known. In this case, we replace cr e by s e and cr~ by s^ in the fore-
going expression. To make a prediction interval for y p , we use the t distribution when a e is not known.
Prediction Interval for y p The (1 - a) 100% prediction interval for the predicted value of y,
denoted by y p , for x = x Q is
y ± %
where the value of t is obtained from the t distribution table for a/2 area in the right tail of the
t distribution curve and df = n — 2. The value of s ? is calculated as follows:
n SS,.
Example 13-10 illustrates the procedure to make a prediction interval for a particular value
of y.
13.9 Cautions in Using Regression 609
■ EXAMPLE 13-10
Refer to Example 13-1 on incomes and food expenditures. Find a 99% prediction interval for
the predicted food expenditure for a randomly selected household with a monthly income of
$5500.
Solution Using the regression line estimated in Example 13—1, we find the point estimate
of the predicted food expenditure for x = 55:
y = 1.5050 + .2525(55) = $15.3925 hundred
The area in each tail of the t distribution for a 99% confidence level is
a/2 = (1 -.99)/2 = .005
The degrees of freedom are
<i/=n-2 = 7- 2 = 5
From the t distribution table, the t value for .005 area in the right tail of the t distribution curve
and 5 dfis 4.032. From calculations in Examples 13-1 and 13-2,
s e = 1.5939, x = 55.1429, and SS AX = 1772.8571
The standard deviation of y as an estimator of y p for x = 55 is calculated as follows:
I - ' • (Vi ' V) '
n SS„.
Making a prediction
interval for a particular
value of y.
1 (55 - 55.1429) 2
(1.5939) J 1 + - + = 1.7040
V ; V 7 1772.8571
Hence, the 99% prediction interval for y p for x = 55 is
y ± ts % = 15.3925 ± 4.032(1.7040)
= 15.3925 ± 6.8705 = 8.5220 to 22.2630
Thus, with 99% confidence we can state that the predicted food expenditure of a house-
hold with a monthly income of $5500 is between $852.20 and $2226.30. ■
As we can observe in Example 13-10, this interval is much wider than the one for the mean
value of y for x = 55 calculated in Example 13-9, which was $1296.32 to $1782.18. This is
always true. The prediction interval for predicting a single value of y is always larger than the
confidence interval for estimating the mean value of y for a certain value of x.
13.9 Cautions in Using Regression
When carefully applied, regression is a very helpful technique for making predictions and esti-
mations about one variable for a certain value of another variable. However, we need to be cau-
tious when using the regression analysis, for it can give us misleading results and predictions.
The following are the two most important points to remember when using regression.
Extrapolation
The regression line estimated for the sample data is reliable only for the range of x values ob-
served in the sample. For example, the values of x in our example on incomes and food ex-
penditures vary from a minimum of 33 to a maximum of 83. Hence, our estimated regression
line is applicable only for values of x between 33 and 83; that is, we should use this regression
line to estimate the mean food expenditure or to predict the food expenditure of a single house-
hold only for income levels between $3300 and $8300. If we estimate or predict y for a value
of x either less than 33 or greater than 83, it is called extrapolation. This does not mean that
we should never use the regression line for extrapolation. Instead, we should interpret such pre-
dictions cautiously and not attach much importance to them.
610 Chapter 13 Simple Linear Regression
Similarly, if the data used for the regression estimation are time-series data (see Exercises
13.100 and 13.101), the predicted values of y for periods outside the time interval used for the
estimation of the regression line should be interpreted very cautiously. When using the estimated
regression line for extrapolation, we are assuming that the same linear relationship between the
two variables holds true for values of x outside the given range. It is possible that the relation-
ship between the two variables may not be linear outside that range. Nonetheless, even if it is
linear, adding a few more observations at either end will probably give a new estimation of the
regression line.
Causality
The regression line does not prove causality between two variables; that is, it does not predict
that a change in y is caused by a change in x. The information about causality is based on the-
ory or common sense. A regression line describes only whether or not a significant quantita-
tive relationship between x and y exists. Significant relationship means that we reject the null
hypothesis H : B = at a given significance level. The estimated regression line gives the
change in y due to a change of one unit in x. Note that it does not indicate that the reason y
has changed is that x has changed. In our example on incomes and food expenditures, it is
economic theory and common sense, not the regression line, that tell us that food expenditure
depends on income. The regression analysis simply helps determine whether or not this de-
pendence is significant.
EXERCISES
CONCEPTS AND PROCEDURES
13.86 Briefly explain the difference between estimating the mean value of y and predicting a particular
value of y using a regression model.
13.87 Construct a 99% confidence interval for the mean value of y and a 99% prediction interval for the
predicted value of y for the following.
a. y = 3.25 + .80* for x = 15 given ^ = .954, x = 18.52, SS„. = 144.65, andn = 10
b. y = -27 + 7.67x for x = 12 given s e = 2.46, x = 13.43, SS^ = 369.77, and n = 10
13.88 Construct a 95% confidence interval for the mean value of y and a 95% prediction interval for the
predicted value of y for the following.
a. y = 13.40 + 2.58xforx = 8 given s e = 1.29, x = 11.30, SS„ = 210.45, andn = 12
b. y = -8.6 + 3.72* for* = 24 given s e = 1.89, x = 19.70, SS^ = 315.40, and n = 10
■ APPLICATIONS
13.89 Refer to Exercise 13.53. Construct a 90% confidence interval for the mean monthly salary of all
secretaries with 10 years of experience. Construct a 90% prediction interval for the monthly salary of a
randomly selected secretary with 10 years of experience.
13.90 Refer to the data on temperature settings and pairs of gloves produced for 8 days given in Exercise
13.80. Construct a 99% confidence interval for (i y i x for x = 77 and a 99% prediction interval for y p for
x = 77.
13.91 Refer to Exercise 13.81. Construct a 95% confidence interval for the mean price of all tires that
have a 65,000-mile limited tread warranty. Construct a 95% prediction interval for the price of a randomly
selected tire that has a 65,000-mile limited tread warranty.
13.92 Refer to Exercise 13.82. Construct a 99% confidence interval for the mean bounce height of all bas-
ketballs that are inflated to 8.5 psi. Construct a 99% prediction interval for the bounce height of a ran-
domly selected basketball that is inflated to 8.5 psi.
13.93 Refer to Exercise 13.83. Construct a 95% confidence interval for the mean charitable contributions
made by all households with an income of $84,000. Make a 95% prediction interval for the charitable con-
tributions made by a randomly selected household with an income of $84,000.
13.94 Refer to Exercise 13.85. Construct a 98% confidence interval for the mean starting salary of recent
college graduates with a GPA of 3.15. Construct a 98% prediction interval for the starting salary of a ran-
domly selected recent college graduate with a GPA of 3.15.
Uses and Misuses 61 1
USES AND MISUSES...
1. PROCESSING ERRORS
Stuck on the far right side of the linear regression model is the Greek
letter epsilon, e. Despite its diminutive size, proper respect for the er-
ror term is critical to good linear regression modeling and analysis.
One interpretation of the error term is that it is a process. Imag-
ine you are a chemist and you have to weigh a number of chemi-
cals for an experiment. The balance that you use in your laboratory
is very accurate— so accurate, in fact, that the shuffling of your feet,
your exhaling near it, or the rumbling of trucks on the road outside
can cause the reading to fluctuate. Because the value of the meas-
urement that you take will be affected by a number of factors out of
your control, you must make several measurements for each chem-
ical, note each measurement, and then take the means and standard
deviations of your samples. The distribution of measurements around
a mean is the result of a random error process dependent on a num-
ber of factors out of your control; each time you use the balance, the
measurement you take is the sum of the actual mass of the chemi-
cal and a "random" error. In this example, the measurements will
most likely be normally distributed around the mean.
Linear regression analysis makes the same assumption about
the two variables you are comparing: The value of the dependent
variable is a linear function of the independent variable, plus a little
bit of error that you cannot control. Unfortunately, when working with
economic or survey data, you rarely can duplicate an experiment to
identify the error model. As a statistician, however, you can use the
errors to help you refine your model of the relationship among the
variables and to guide your collection of new data. For example, if
the errors are skewed to the right for moderate values of the inde-
pendent variable and skewed to the left for small and large values
of the independent variable, you can modify your model to account
for this difference. Or you can think about other relationships among
the variables that might explain this particular distribution of errors.
A detailed analysis of the error in your model can be just as instruc-
tive as analysis of the slope and y-intercept of the identified model.
2. OUTLIERS AND CORRELATION
In Chapter 3 we learned that outliers can affect the values of some
of the summary measures such as the mean and range. Note that
although outliers do affect many other summary measures, these
two are affected substantially. Here we will see that just looking at
a number that represents the correlation coefficient does not pro-
vide the entire story. A very famous data set for demonstrating this
concept was created by F. J. Anscombe (Anscombe, F. J., Graphs in
Statistical Analysis, American Statistician, 27, pp. 17-21). He created
four pairs of data sets on x and y variables, each of which has a
correlation of .816. To the novice, it may seem that the scatterplots
for these four data sets should look virtually the same, but that may
not be true. Look at the four scatterplots shown in Figure 13.25.
No two of these scatterplots are even remotely close to being
the same or even similar. The data used in the upper left plot are
linearly associated, as are the data in the lower left plot. However
the plot of y3 versus x3 contains an outlier. Without this outlier,
the correlation between x3 and y3 would be 1. On the other hand,
there is much more variability in the relationship between xl and
10
8
6
4
12.5
10.0
7.5
5.0
Scatterplots of Anscombe 's data
yl*xl
10
y2*x2
5.0 7.5 10.0 12.5
15.0
5.0
7,5 10.0
12,5 15.0
y3*x3
y4*x4
•
12,5-
10,0-
•
•
7,5-
•
•
•
1
s.n -
!
5.0 7.5 10.0 12.5 15,0
10
15
20
Figure T3.25 Four scatterplots with the same correlation coefficient.
612 Chapter 13 Simple Linear Regression
y\. As far asx4 and y4 are concerned, the strong correlation is de-
fined by the single point in the upper right corner of the scatter-
plot. Without this point, there would be no variability among the
x4 values, and the correlation would be undefined. Lastly, the
scatterplot of yl versus x2 reveals that there is an extremely well-
defined relationship between these variables, but it is not linear.
Being satisfied that the correlation coefficient is close to 1.0 be-
tween variables x2 and yl implies that there is a strong linear as-
sociation between the variables when actually we are fitting a line
to a set of data that should be represented by another type of
mathematical function.
As we have mentioned before, the process of making a graph
may seem trivial, but the importance of graphs in our analysis can
never be overstated.
Glossary
Coefficient of determination A measure that gives the proportion
(or percentage) of the total variation in a dependent variable that is
explained by a given independent variable.
Degrees of freedom for a simple linear regression model Sam-
ple size minus 2; that is, n — 2.
Dependent variable The variable to be predicted or explained.
Deterministic model A model in which the independent variable
determines the dependent variable exactly. Such a model gives an
exact relationship between two variables.
Estimated or predicted value of y The value of the dependent
variable, denoted by y, that is calculated for a given value of x us-
ing the estimated regression model.
Independent or explanatory variable The variable included in a
model to explain the variation in the dependent variable.
Least squares estimates of A and B The values of a and b that
are calculated by using the sample data.
Least squares method The method used to fit a regression line
through a scatter diagram such that the error sum of squares is
minimum.
Least squares regression line A regression line obtained by us-
ing the least squares method.
Linear correlation coefficient A measure of the strength of the
linear relationship between two variables.
Linear regression model A regression model that gives a straight-
line relationship between two variables.
Multiple regression model A regression model that contains two
or more independent variables.
Negative relationship between two variables The value of the
slope in the regression line and the correlation coefficient between
two variables are both negative.
Nonlinear (simple) regression model A regression model that does
not give a straight-line relationship between two variables.
Population parameters for a simple regression model The val-
ues of A and B for the regression model y = A + Bx + e that are
obtained by using population data.
Positive relationship between two variables The value of the
slope in the regression line and the correlation coefficient between
two variables are both positive.
Prediction interval The confidence interval for a particular value
of y for a given value of x.
Probabilistic or statistical model A model in which the independent
variable does not determine the dependent variable exactly.
Random error term (e) The difference between the actual and
predicted values of y.
Scatter diagram or scatterplot A plot of the paired observations
of x and y.
Simple linear regression A regression model with one dependent
and one independent variable that assumes a straight-line relation-
ship.
Slope The coefficient of x in a regression model that gives the
change in y for a change of one unit in x.
SSE (error sum of squares) The sum of the squared differences
between the actual and predicted values of y. It is the portion of
the SST that is not explained by the regression model.
SSR (regression sum of squares) The portion of the SST that is
explained by the regression model.
SST (total sum of squares) The sum of the squared differences be-
tween actual y values and y.
Standard deviation of errors A measure of spread for the ran-
dom errors.
y-intercept The point at which the regression line intersects the
vertical axis on which the dependent variable is marked. It is the
value of y when x is zero.
Supplementary Exercises
13.95 The following data give information on the ages (in years) and the numbers of breakdowns during
the last month for a sample of seven machines at a large company.
Age (years)
12
7
2
8
13
9
4
Number of breakdowns
10
5
1
4
12
7
2
Supplementary Exercises 613
a. Taking age as an independent variable and number of breakdowns as a dependent variable, what
is your hypothesis about the sign of B in the regression line? (In other words, do you expect B
to be positive or negative?)
b. Find the least squares regression line. Is the sign of b the same as you hypothesized for B in part a?
c. Give a brief interpretation of the values of a and b calculated in part b.
d. Compute r and r and explain what they mean.
e. Compute the standard deviation of errors.
f. Construct a 99% confidence interval for B.
g. Test at the 2.5% significance level whether B is positive.
h. At the 2.5% significance level, can you conclude that p is positive? Is your conclusion the same
as in part g?
13.96 The health department of a large city has developed an air pollution index that measures the level of
several air pollutants that cause respiratory distress in humans. The accompanying table gives the pollution
index (on a scale of 1 to 10, with 10 being the worst) for 7 randomly selected summer days and the num-
ber of patients with acute respiratory problems admitted to the emergency rooms of the city's hospitals.
Air pollution index
4.5
6.7
8.2
5.0
4.6
6.1
3.0
Emergency admissions
53
82
102
60
39
42
27
a. Taking the air pollution index as an independent variable and the number of emergency admis-
sions as a dependent variable, do you expect B to be positive or negative in the regression model
y = A + Bx + e?
b. Find the least squares regression line. Is the sign of b the same as you hypothesized for B in part a?
c. Compute r and r 2 , and explain what they mean.
d. Compute the standard deviation of errors.
e. Construct a 90% confidence interval for B.
f. Test at the 5% significance level whether B is positive.
g. Test at the 5% significance level whether p is positive. Is your conclusion the same as in part f?
13.97 The management of a supermarket wants to find if there is a relationship between the number of
times a specific product is promoted on the intercom system in the store and the number of units of that
product sold. To experiment, the management selected a product and promoted it on the intercom system
for 7 days. The following table gives the number of times this product was promoted each day and the
number of units sold.
Number of Promotions
per Day
Number of Units Sold
per Day (hundreds)
15
11
22
22
42
30
30
26
18
17
12
15
38
23
a. With the number of promotions as an independent variable and the number of units sold as a de-
pendent variable, what do you expect the sign of B in the regression line y = A + Bx + e will be?
b. Find the least squares regression line y = a + bx. Is the sign of b the same as you hypothesized
for B in part a?
c. Give a brief interpretation of the values of a and b calculated in part b.
d. Compute r and r and explain what they mean.
e. Predict the number of units of this product sold on a day with 35 promotions.
f. Compute the standard deviation of errors.
g. Construct a 98% confidence interval for B.
h. Testing at the 1% significance level, can you conclude that B is positive?
i. Using a = .02, can you conclude that the correlation coefficient is different from zero?
13.98 The following table provides information on the high temperature (nws.noaa.gov) for each day and
the number of crimes committed in Chicago, Illinois, during the period July 1, 2009 to July 14, 2009
(http://chicago.everyblock.com/crime/).
614 Chapter 13 Simple Linear Regression
High temperature (°F)
65
73
79
69
81
86
77
Number of crimes
1110
1134
1117
1044
1014
1105
1152
High temperature (°F)
65
79
82
85
82
79
80
Number of crimes
1046
1127
1160
1065
1126
1041
1038
a. Find the least squares regression line y = a + bx. Take high temperature as an independent vari-
able and number of crimes committed as a dependent variable.
b. Give a brief interpretation of the values of a and b.
c. Compute r and r and explain what they mean.
d. Predict the number of crimes committed on a day with a high temperature of 83°F
e. Compute the standard deviation of errors.
f. Construct a 99% confidence interval for B.
g. Testing at the 1% significance level, can you conclude that B is different from zero?
h. Using a =.01, can you conclude that the correlation coefficient is different from zero?
13.99 The following table provides information on the 10 NASDAQ companies with the largest percent-
age of their stocks traded on July 6, 2009. Specifically, the table gives the information on the percentage
of stocks traded and the change (in dollars per share) in each stock's price.
Stock
Percentage Traded
Change ($)
Matrixx
19.6
-0.43
SpectPh
14.7
-1.10
DataDom
12.4
0.85
CardioNet
9.3
-0.11
DryShips
9.0
-0.42
DynMatl
8.0
-1.16
EvrgrSlr
7.9
-0.04
EagleBulk
7.9
-0.11
Palm
7.8
-0.02
CentAl
7.4
-0.57
Source: USA TODAY, July 7, 2009.
a. With percentage traded as an independent variable and the change in the stock's price as a de-
pendent variable, compute SS KX , SS yy , and SS yv .
b. Construct a scatter diagram for these data. Does the scatter diagram exhibit a negative linear re-
lationship between the percentage of stock traded and the change in the stock's price?
c. Find the regression equation y = a + bx.
d. Give a brief interpretation of the values a and b calculated in part c.
e. Compute the correlation coefficient r.
f. Predict the change in a stock's price if 8.6% of the stock's shares are traded on a day. Using part b,
how reliable do you think this prediction will be? Explain.
13.100 The following table gives the total daily U.S. crude oil imports (in millions of barrels, rounded to
the nearest million) for the years 1995 to 2008.
Year
1995
1996
1997
1998
1999
2000
2001
Daily U.S. crude oil imports
(millions of barrels)
7.23
7.51
8.23
8.71
8.73
9.07
9.33
Year
2002
2003
2004
2005
2006
2007
2008
Daily U.S. crude oil imports
(millions of barrels)
9.14
9.66
10.08
10.13
10.12
10.03
9.78
Source: http://tonto.eia.doe.gov/dnav/pet/hist/mcrimus2a.htm.
a. Assign a value of to 1995, 1 to 1996, 2 to 1997, and so on. Call this new variable Time. Make
a new table with the variables Time and Daily U.S. Crude Oil Imports.
Supplementary Exercises 615
b. With time as an independent variable and the daily U.S. crude oil imports as the dependent vari-
able, compute SS XX , SS yy , and SS xy .
c. Construct a scatter diagram for these data. Does the scatter diagram exhibit a linear positive re-
lationship between time and daily U.S. crude oil imports?
d. Find the least squares regression line y = a + bx.
e. Give a brief interpretation of the values of a and b calculated in part d.
f. Compute the correlation coefficient r.
g. Predict the daily U.S. crude oil imports for x = 20. Comment on this prediction.
h. Recalculate the correlation coefficient, ignoring the data for 2006, 2007, and 2008. What hap-
pens to the value of the correlation coefficient? Create a scatter diagram of the data with time
on the horizontal axis and imports on the vertical axis. Use the diagram to explain what hap-
pened to the value of r.
13.101 The following table gives the completion times for the winners in the women's 200-meter dash fi-
nals in the Summer Olympic Games from 1972 to 2008. The times are in seconds rounded to the nearest
1/100 second.
Olympic Year
Time (seconds)
1972
22.40
1976
22.37
1980
22.03
1984
21.81
1988
21.34
1992
21.81
1996
22.12
2000
21.85
2004
22.05
2008
21.74
Source: Wikipedia.
a. Assign a value of to 1972, 1 to 1976, 2 to 1980, and so on. Call this new variable Year. Make
a new table with the variables Year and Time.
b. With year as an independent variable and time as the dependent variable, compute SS a , S5 VV ,
and SS^
c. Construct a scatter diagram for these data. Does the scatter diagram exhibit a linear negative
relationship between year and time?
d. Find the least squares regression line y = a + bx.
e. Give a brief interpretation of the values of a and b calculated in part d.
f. Compute the correlation coefficient r.
g. Predict the time for the year 2016. Comment on this prediction.
13.102 Refer to the data on ages and numbers of breakdowns for seven machines given in Exercise 13.95.
Construct a 99% confidence interval for the mean number of breakdowns per month for all machines with
an age of 8 years. Find a 99% prediction interval for the number of breakdowns per month for a randomly
selected machine with an age of 8 years.
13.103 Refer to the data on the air pollution index and the number of emergency hospital admissions for
acute respiratory problems given in Exercise 13.96. Determine a 95% confidence interval for the mean
number of such emergency admissions on all days with an air pollution index of 7.0. Make a 95%
prediction interval for the number of such emergency admissions on a day when the air pollution index
is 7.0.
13.104 Refer to the data given in Exercise 13.97 on the number of times a specific product is promoted
on the intercom system in a supermarket and the number of units of that product sold. Make a 90%
confidence interval for the mean number of units of that product sold on days with 35 promotions. Con-
struct a 90% prediction interval for the number of units of that product sold on a randomly selected day
with 35 promotions.
13.105 Refer to the data given in Exercise 13.98 on the high temperature on each day and the numbers
of crimes committed in Chicago for a period of 14 days. Construct a 98% confidence interval for the mean
number of crimes committed on all days with the high temperature of 85°F Construct a 98% prediction
616 Chapter 13 Simple Linear Regression
interval for the number of crimes committed on a randomly chosen day when the highest temperature
is 85°F.
Advanced Exercises
13.106 Consider the data given in the following table.
X
10
20
30
40
50
60
y
12
15
19
21
25
30
a. Find the least squares regression line and the linear correlation coefficient r.
b. Suppose that each value of y given in the table is increased by 5 and the x values remain un-
changed. Would you expect r to increase, decrease, or remain the same? How do you expect
the least squares regression line to change?
c. Increase each value of y given in the table by 5 and find the new least squares regres-
sion line and the correlation coefficient r. Do these results agree with your expectation
in part b?
13.107 Suppose that you work part-time at a bowling alley that is open daily from noon to midnight.
Although business is usually slow from noon to 6 p.m., the owner has noticed that it is better on hotter
days during the summer, perhaps because the premises are comfortably air-conditioned. The owner
shows you some data that she gathered last summer. This data set includes the maximum temperature
and the number of lines bowled between noon and 6 p.m. for each of 20 days. (The maximum
temperatures ranged from 77°F to 95°F during this period.) The owner would like to know if she can
estimate tomorrow's business from noon to 6 p.m. by looking at tomorrow's weather forecast. She asks
you to analyze the data. Let x be the maximum temperature for a day and y the number of lines bowled
between noon and 6 p.m. on that day. The computer output based on the data for 20 days provided the
following results:
y = -432 + 7.7*, s e = 28.17, SS n . = 607, and x = 87.5
Assume that the weather forecasts are reasonably accurate.
a. Does the maximum temperature seem to be a useful predictor of bowling activity between
noon and 6 p.m.? Use an appropriate statistical procedure based on the information given. Use
a = .05.
b. The owner wants to know how many lines of bowling she can expect, on average, for days with
a maximum temperature of 90°. Answer using a 95% confidence level.
c. The owner has seen tomorrow's weather forecast, which predicts a high of 90°F About how
many lines of bowling can she expect? Answer using a 95% confidence level.
d. Give a brief commonsense explanation to the owner for the difference in the interval estimates
of parts b and c.
e. The owner asks you how many lines of bowling she could expect if the high temperature were
100°F Give a point estimate, together with an appropriate warning to the owner.
13.108 An economist is studying the relationship between the incomes of fathers and their sons or
daughters. Let x be the annual income of a 30-year-old person and let y be the annual income of that
person's father at age 30 years, adjusted for inflation. A random sample of 300 thirty-year-olds and their
fathers yields a linear correlation coefficient of .60 between x and y. A friend of yours, who has read
about this research, asks you several questions, such as: Does the positive value of the correlation
coefficient suggest that the 30-year-olds tend to earn more than their fathers? Does the correlation
coefficient reveal anything at all about the difference between the incomes of 30-year-olds and their
fathers? If not, what other information would we need from this study? What does the correlation
coefficient tell us about the relationship between the two variables in this example? Write a short note
to your friend answering these questions.
13.109 For the past 25 years Burton Hodge has been keeping track of how many times he mows his
lawn and the average size of the ears of corn in his garden. Hearing about the Pearson correlation co-
efficient from a statistician buddy of his. Burton decides to substantiate his suspicion that the more of-
ten he mows his lawn, the bigger are the ears of corn. He does so by computing the correlation coeffi-
cient. Lo and behold, Burton finds a .93 coefficient of correlation! Elated, he calls his friend the
statistician to thank him and announce that next year he will have prize-winning ears of corn because
he plans to mow his lawn every day. Do you think Burton's logic is correct? If not, how would you
explain to Burton the mistake he is making in his presumption (without eroding his new opinion of
Self-Review Test 617
statistics)? Suggest what Burton could do next year to make the ears of corn large, and relate this to the
Pearson correlation coefficient.
13.110 It seems reasonable that the more hours per week a full-time college student works at a job, the
less time he or she will have to study and, consequently, the lower his or her GPA would be.
a. Assuming a linear relationship, suggest specifically what the equation relating x and y would
be, where x is the average number of hours a student works per week and y represents the stu-
dent's GPA. Try several values of x and see if your equation gives reasonable values of y.
b. Using the following observations taken from 10 randomly selected students, compute the re-
gression equation and compare it to yours of part a.
Average number
of hours worked
20
28
10
35
5
14
40
8
23
GPA
2.8
2.5
3.1
2.1
3.4
3.3
2.8
2.5
3.6
1.8
13.111 Consider the formulas for calculating a prediction interval for a new (specific) value of y. For each
of the changes mentioned in parts a through c below, state the effect on the width of the confidence in-
terval (increase, decrease, or no change) and why it happens. Note that besides the change mentioned in
each part, everything else such as the values of a, b, x, s e , and 55 vr remains unchanged.
a. The confidence level is increased.
b. The sample size is increased.
c. The value of x is moved farther away from the value of x.
d. What will the value of the margin of error be if x equals jc?
13.112 For each of the regression lines in Exercises 13.53 through 13.56, interpret the slope in terms of
the application of that exercise. In addition, state whether the value of the intercept is logical, and why it
is or is not logical. If it is logical, state what the value of the intercept represents in terms of the specific
application of that exercise.
13.113 Consider the following data
X
-5
-4
-3
-2
-1
1
2
3
4
5
y
-125
-64
-27
-8
-1
1
8
27
64
125
a. Calculate the correlation between x and y, and perform a hypothesis test to determine if the
correlation is significantly greater than zero. Use a significance level of 5%.
b. Are you willing to conclude that there is a strong linear association between the two variables?
Use at least one graph to support your answer, and to explain why or why not.
Self-Review Test
1. A simple regression is a regression model that contains
a. only one independent variable
b. only one dependent variable
c. more than one independent variable
d. both a and b
2. The relationship between independent and dependent variables represented by the (simple) linear re-
gression is that of
a. a straight line b. a curve c. both a and b
3. A deterministic regression model is a model that
a. contains the random error term
b. does not contain the random error term
c. gives a nonlinear relationship
4. A probabilistic regression model is a model that
a. contains the random error term
b. does not contain the random error term
c. shows an exact relationship
618 Chapter 13 Simple Linear Regression
5. The least squares regression line minimizes the sum of
a. errors b. squared errors c. predictions
6. The degrees of freedom for a simple regression model are
a. n — 1 b. n — 2 c. n — 5
7. Is the following statement true or false?
The coefficient of determination gives the proportion of total squared errors (SST) that is
explained by the use of the regression model.
8. Is the following statement true or false?
The linear correlation coefficient measures the strength of the linear association between two
variables.
9. The value of the coefficient of determination is always in the range
a. to 1 b. - 1 to 1 c. - 1 to
10. The value of the correlation coefficient is always in the range
a. to 1 b. - 1 to 1 c. - 1 to
11. Explain why the random error term e is added to the regression model.
12. Explain the difference between A and a and between B and b for a regression model.
13. Briefly explain the assumptions of a regression model.
14. Briefly explain the difference between the population regression line and a sample regression line.
15. The following table gives the temperatures (in degrees Fahrenheit) at 6 p.m. and the attendance
(rounded to hundreds) at a minor league baseball team's night games on 7 randomly selected evenings
in May.
Temperature
61
70
50
65
48
75
53
Attendance
10
16
12
15
8
20
18
a. Do you think temperature depends on attendance or attendance depends on temperature?
b. With temperature as an independent variable and attendance as a dependent variable, what is your
hypothesis about the sign of B in the regression model?
c. Construct a scatter diagram for these data. Does the scatter diagram exhibit a linear relationship
between the two variables?
d. Find the least squares regression line. Is the sign of b the same as the one you hypothesized for
B in part b?
e. Give a brief interpretation of the values of the y-intercept and slope calculated in part d.
f. Compute r and r 2 , and explain what they mean.
g. Predict the attendance at a night game in May for a temperature of 60°F
h. Compute the standard deviation of errors.
i. Construct a 99% confidence interval for B.
j. Testing at the 1% significance level, can you conclude that B is positive?
k. Construct a 95% confidence interval for the mean attendance at a night game in May when the
temperature is 60°F.
1. Make a 95% prediction interval for the attendance at a night game in May when the temperature
is 60°F
m. Using the 1% significance level, can you conclude that the linear correlation coefficient is
positive?
Mini-Projects
■ MINI-PROJECT 13-1
Using the weather sections from back issues of a local newspaper or some other source, do the following
for a period of 30 or more days. For each day, record the predicted maximum temperature for the next
Mini-Projects 619
day, and then find the actual maximum temperature in the next day's newspaper. Thus, you will have the
predicted and actual maximum temperatures for 30 or more days.
a. Make a scatter diagram for your data.
b. Find the regression line with actual maximum temperature as a dependent variable and predicted
maximum temperature as an independent variable.
c. Using the 1% significance level, can you conclude that the slope of the regression line is different
from zero?
d. If the actual maximum temperatures were exactly the same as the predicted maximum tempera-
tures for all days, what would the value of the correlation coefficient be?
e. Find the correlation coefficient between the predicted and actual maximum temperatures for your
data.
f. Using the 1% significance level, can you conclude that the linear correlation coefficient is
positive?
■ MINI-PROJECT 13-2
Two friends are arguing about the relationship between the prices of soft drinks and wine in U.S. cities.
Justin thinks that the prices of any two types of beverages (a soft drink and wine) should be positively re-
lated. Ivan disagrees, arguing that the prices of alcoholic beverages in a city depend primarily on state and
local taxes.
a. Take a random sample of 15 U.S. cities from the City Data that accompany this text. Let x
be the price of a 2-liter bottle of Coca-Cola and y the price of a 1.5-liter bottle of Livingston
Cellars or Gallo Chablis or Chenin Blanc wine. Calculate the linear correlation coefficient
between x and y.
b. Does your value of r suggest a positive linear relationship between x and y?
c. Do you think finding a regression line makes sense here?
d. Using the 1% level of significance, can you conclude that the linear correlation coefficient is
positive?
■ MINI-PROJECT 13-3
Visit a grocery store and choose 30 different types of food items that include nutrition information on
the packaging. For each food, identify the amount of fat (in grams) and the sodium content (in milligrams)
per serving. Make sure that you pick a wide variety of foods in order to get a wide variety of values of
these two variables. For example, selecting 30 different diet sodas would not make for an interesting
analysis.
a. Calculate the linear correlation coefficient between the two variables. Do you find a positive or a
negative association between sodium content and the amount of fat?
b. Create a scatterplot of these data using the amount of fat as the x variable. Does your scatterplot
suggest that creating a regression line to represent these data makes sense?
c. Find a regression line for your data. If it makes sense to fit a line, interpret the values of the
slope and intercept. If it does not make sense, explain why these numbers could be
misleading.
■ MINI-PROJECT 13-4
Using Data Set VIII (McDonald's Data) that is on the Web site of this text, take a random sample of 12
items each from the list of sandwiches, non-sandwich chicken products, and breakfast entrees. Record the
number of grams of carbohydrates and the number of calories.
a. Calculate the correlation coefficient between the two variables. Do you find a positive or a nega-
tive association between carbohydrate content and the number of calories?
b. Create a scatterplot of these data using the carbohydrate content as the x variable. Does your scat-
terplot suggest that fitting a regression line to these data makes sense?
c. Find the equation of the estimated regression line for your data. If it makes sense to fit a line, in-
terpret the values of the slope and intercept. If it does not make sense, explain why the numbers
could be misleading.
620 Chapter 13 Simple Linear Regression
DECIDE FOR YOURSELF
Does Regression Equation Always
Make Sense?
Regression is a very powerful statistical tool. However, like any
other tool, a failure to understand both its uses as well as its limita-
tions can lead to ridiculous, if not disastrous, results. To demon-
strate this, we took the data on two variables — the year of the
Olympics from 1928 to 2004 as the independent variable and winner's
time (in seconds) in the men's 100 meter dash (race) as a dependent
variable. Figure 13.26 shows the scatterplot and the regression line
for these data.
Statrerplot of Seconds vsVear
10 e ■ t
1S20 WOO l*tO 1560 1560 1570 1980 1990 20DO 20 10
Year
Figure 13.26 Scatterplot and regression line.
Looking at this scatterplot, it seems reasonable to use a regres-
sion line to explain the relationship between the year of Olympics
and the winning time in the 100 meter dash. Specifically, the equa-
tion of that regression line is
Seconds = 31.1 - .0106 Year
To calculate this regression line, we used the actual years of the
Olympiad for the independent variable. Theoretically, we could use
this regression equation to estimate the winning times for the years
when Olympics are not held. We could also use it to predict the
future winning times or to calculate what would have happened in
the past. Answer the following questions to see how reasonable this
process is.
1. Based on this regression equation, what is the change in the win-
ning time per Olympic period (4 years)? Does the change represent
an increase or a decrease?
2. Find the predicted winning times for the years 2200, 2600, and
3000. Using these predicted times, determine the winners' speeds
(in miles per hour) for the years 2200 and 2600. Does it make
sense to believe that this pattern will continue in the future?
Explain.
3. A similar analysis could be done in the reverse direction. A 2005
scientific discovery stated that fossils from 35,000-year-old modern
humans were found in Transylvania (http://www.theglobeandmail
.com/servlet/story/RTGAM.20040306.wfossil0306/BNStory/special
ScienceandHealth/). Using the above regression equation, calculate
the winning time for the 100 meter dash at this point in history. Does
this number make sense? Why or why not?
ECHNOLOGY
INSTRUCTION
Simple Linear Regression
L i riRe*3 < a+bx ) Li 7
Ls,Vii
Screen 13.1
1. To construct a simple linear regression equation, enter the independent and dependent
variable values into lists. Select STAT >CALC >LinReg(a+bx), press ENTER, then
enter (separated by commas) the name of your independent variable list, the name of your
dependent variable list, and Yl. (Yl can be found by selecting VARS >Y-VARS >
Function >Y1.) Then press ENTER. (See Screen 13.1.) The result includes the slope
and intercept of the regression equation.
2. To find the correlation coefficient, select VARS >Statistics >EQ >r. To find the coeffi-
cient of determination, square the correlation coefficient.
3. To find a fitted value for a given value of x, type Yl(x).
4. To test that the slope of the line is nonzero, select STATS >TESTS >LinRegTTest.
(Note that this set of commands will give you the output obtained under 1 and 2 above.)
Enter the names of the lists. Leave Freq:l. Choose the alternative hypothesis. Select Cal-
culate. The result includes a f-statistic value and a ;?-value.
Technology Instruction 621
Regression
CI
C2
Response: [y~
Predictors:
Select
Help
Screen 13.2
1. To construct and analyze a simple linear re-
gression equation, enter the independent and
dependent variable values into columns.
2. Select Stat >Regression >Regression.
3. Enter the dependent variable's column
name in the Response box.
4. Enter the independent variable's column
name in the Predictors box. (See Screen
13.2.)
5. Select Options if you wish to predict a
value with the equation, and enter the
value of the independent variable in the en-
try marked Prediction intervals for new
observations. Enter the Confidence level
and select OK.
6. Select Results, and choose Regression
equation, .... Select OK for each dialog
box.
7. The output includes the regression equation, t statistics and p- values for tests on both the slope
and intercept to find out if they are zero, the coefficient of determination (as R-sq), and, if re-
quested, the fitted value, as well as confidence and prediction intervals for the fitted value.
Graphs.
Results.
Options...
Storage...
OK
Cancel
ID
11
12
A B
Fertilizer Yield
120
SO
100
70
SB
75
110
Correlation
Screen 13.3
142|
111
132|
96
us;
104j
136!
mm
Input
Input Range:
Grouped By:
W\ Labels in First Row
Output options
O Output Range:
© New Worksheet Ply:
O New Workbook
m
OK
Columns
O Rows
Cancel
Help
A
B
C
1
Fertilizer
Yield
2
Fertilizer
1
3
Yield
0.981432
1
Screen 13.4
Click the Data tab. Click the Data
Analysis button within the Analy-
sis group.
To calculate the linear correlation
coefficient, select Correlation. En-
ter the location of the data in the
Input Range box. Click the button
to identify whether the data for
each sample are given in columns
or rows. If your data have labels in
the top row (or in the left column),
click the Labels box. Choose how
you wish the output to appear. (See
Screen 13.3.) Click OK.
3. The output contains a two-by-two
table. The value in the lower left is
the value of the correlation between
the two variables. (See Screen 13.4.)
To calculate the coefficients of the least squares regression line, perform a hypothesis
test on the slope of the regression line, and calculate a confidence interval for the slope
of the regression line, select Regression from the list of choices within the Data
622 Chapter 13 Simple Linear Regression
1
Fpr"t"i ii7pr
rci li 1 1 1, c I
Yield
2
120
142
3
SO
112
4
100
132
5
70
95
5
83
119
7
75
104
g
11C
135
g
10
11
12
13
14
13
15
17
18
10
Screen 13.5
Regression
OK
|A$l:fAf8
Cancel
I I Constant is Zero
95
Help
Analysis dialog box. Enter the
location of the data in the Input
Range box. Click the button to
identify whether the data for
each sample are given in
columns or rows. If your data
have labels in the top row (or in
the left column), click the Labels
box. Enter the confidence level
if you want something other
than 95% confidence interval.
Choose how you wish the
output to appear. (See Screen
13.5.) Click OK.
, The output contains three tables.
The first table, labeled Regres-
sion Statistics, contains the
standard deviation of errors in
the line labeled Standard
Error. In the bottom table, the
Coefficients column contains
the values of a and b. The
remaining values in the top row of this table are not used with respect to this book. The re-
maining descriptions correspond to the values in the bottom row of the bottom table. The
value in the Standard Error column is the value of s b . The next two columns contain the
value of the test statistic and the two-sided /7-value for a test with the slope coefficient equal
to zero. The next two columns give the endpoints of the 95% confidence interval for B. If
you requested a confidence level other than 95%, the endpoints will be in the last two
columns. (See Screen 13.6.)
Input
Input Y Range:
Input X Range:
Labels
1 I Confidence Level:
Output options
O Output Range:
New Worksheet Ply:
New Workbook
Residuals
1 I Residuals
I I Standardized Residuals
Normal Probability
□ Normal Probability Plots
□ Residual Plots
□ Line Fit Plots
1
A
B
E
F
H
1
SUMMARY OUTPUT
2
3
Regression Statistics
4
Multiple R
0.981432419
5
RSquare
0.963209592
e
Adjusted R Square
0.955851511
7
Standard Error
3.619902228
8
Observations
7
9
10
ANOVA
11
*
MS
F
Significance F
12
Regression
1
1715.338682
1715.339
130.90499
8.9 3 3 23 E- 05
13
Residual
5
55.5134507
13.10369
14
Total
6
1780.857143
15
16
Coefficients
Standard Error
tStat
P-value
Lower 95%
Upper 95%
Lower 95.0%
Upper 95.0%
17
Intercept
37.2195602
7.375632241
5.046253
0.0039453
18.25976541
56.179355
18.25976541
56.1793.5499
13
Fertilizer
0.902741957
0.078901549
11.44137
8.933E-05
0.69991907
1.10556485
0.699919C7
1.105554845
Screen 13.6
TECHNOLOGY ASSIGNMENTS
Technology Assignments 623
TA 13.1 In a rainy coastal town in the Pacific Northwest, the local TV weatherman is often criticized
for making inaccurate forecasts for daily precipitation. On each of 30 randomly selected days last winter,
his precipitation forecast (x) for the next day was recorded along with the actual precipitation (y) for that
day. These data are shown in the following table (in inches of rain).
X
y
X
y
X
y
1.0
.6
A
.2
.1
.i
.2
.5
.2
.1
.2
.1
.1
.2
.2
.2
.5
.3
.1
.1
1.0
1.4
2.0
2.1
.2
.1
.5
.3
.4
.2
1.4
1.2
.1
.1
.2
.1
.5
1.0
.1
.5
2.0
.3
.3
.2
Do the following.
a. Construct a scatter diagram for these data.
b. Find the correlation coefficient between the two variables.
c. Find the regression line with actual precipitation as a dependent variable and predicted precipitation
as an independent variable.
d. Make a 95% confidence interval for B.
e. Test at the 1% significance level whether B is positive.
f. Using the 1% significance level, can you conclude that the linear correlation coefficient is positive?
TA13.2 Refer to Data Set III on NBA players. Select a random sample of 30 players from that popula-
tion. Do the following for the data on heights and weights of these 30 players.
a. Construct a scatter diagram for these data.
b. Find the correlation between these two variables.
c. Find the regression line with weight as a dependent variable and height as an independent variable.
d. Make a 90% confidence interval for B.
e. Test at the 5% significance level whether B is positive.
f. Make a 95% confidence interval for the mean weight of all NBA players who are 78 inches tall. Con-
struct a 95% prediction interval for the weight of a randomly selected NBA player with a height of 78
inches.
TA13.3 Refer to the data on the ages and the numbers of breakdowns for a sample of seven machines
given in Exercise 13.95. Answer the following questions.
a. Construct a scatter diagram for these data.
b. Find the least squares regression line with age as an independent variable and the number of break-
downs as a dependent variable.
c. Compute the correlation coefficient.
d. Construct a 99% confidence interval for B.
e. Test at the 2.5% significance level whether B is positive.
Chapter
Multiple Regression
14.1 Multiple Regression
Analysis
14.2 Assumptions of the
Multiple Regression
Model
14.3 Standard Deviation
of Errors
14.4 Coefficient of Multiple
Determination
14.5 Computer Solution of
Multiple Regression
This chapter is not included in this text but is available for download on the
Web site at www.wiley.com/college/mann.
624
Chapter
Nonparametric Methods
This chapter is not included in this text but is available for download on the
Web site at www.wiley.com/college/mann.
15.1 The Sign Test
15.2 The Wilcoxon Signed-
Rank Test for Two
Dependent Samples
1 5.3 The Wilcoxon Rank Sum
Test for Two
Independent Samples
15.4 The Kruskal-Wallis Test
15.5 The Spearman Rho Rank
Correlation Coefficient
Test
15.6 The Runs Test for
Randomness
625
Appendix
Sample Surveys, Sampling Techniques,
and Design of Experiments
The current American fear of germs is evident in the booming sales of antibacterial soaps. They
now represent a large share of the liquid soap market, in spite of a lack of evidence that they are
any better than regular soaps. Do these antibacterial soaps work, or is it just a fad? Are people
using them only because it is their perception that they work to kill germs, or do they really work?
(See Case Study A-2.) Proper sampling techniques employed in research studies or surveys can
help accurately answer questions like these.
A.1 Sources of Data
Case Study A-1 Is It a Simple
Question?
A.2 Sample Surveys and
Sampling Techniques
A.3 Design of Experiments
Case Study A-2 Do
Antibacterial Soaps Work?
A.1 Sources of Data
The availability of accurate data is essential for deriving reliable results and making accurate
decisions. As the truism "garbage in, garbage out" (GIGO) indicates, policy decisions based on
the results of poor data may prove to be disastrous.
Data sources can be divided into three categories: internal sources, external sources, and
surveys and experiments.
A.l.l Internal Sources
Often data come from internal sources, such as a company's personnel files or accounting
records. A company that wants to forecast the future sales of its products might use data from
its records for previous periods. A police department might use data that exist in its records to
analyze changes in the nature of crimes over a period of time.
A2 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
A.1.2 External Sources
All needed data may not be available from internal sources. Hence, to obtain data we may have
to depend on sources outside the company, called external sources. Data obtained from exter-
nal sources may be primary or secondary data. Data obtained from the organization that origi-
nally collected them are called primary data. If we obtain data from the Bureau of Labor Sta-
tistics that were collected by this organization, then these are primary data. Data obtained from
a source that did not originally collect them are called secondary data. For example, data orig-
inally collected by the Bureau of Labor Statistics and published in the Statistical Abstract of
the United States are secondary data.
A.I. 3 Surveys and Experiments
Sometimes the data we need may not be available from internal or external sources. In such
cases, we may have to obtain data by conducting our own survey or experiment.
Surveys
In a survey, we do not exercise any control over the factors when we collect information.
Definition
Survey In a survey, data are collected from the members of a population or sample in such a
way that we have no particular control over the factors that may affect the characteristic of in-
terest or the results of the survey.
For example, if we want to collect data on the money various families spent last month on
clothes, we will ask each of the families included in the survey how much it spent last month
on clothes. Then we will record this information.
A survey may be a census or a sample survey.
(i) Census
A census includes every member of the population of interest, which is called the target pop-
ulation.
Definition
Census A survey that includes every member of the population is called a census.
In practice, a census is rarely taken because it is very expensive and time consuming. Fur-
thermore, in many cases it is impossible to identify each member of the target population. We
discuss these reasons in more detail in Section A. 2.1.
(ii) Sample Survey
Usually, to conduct research, we select a portion of the target population. This portion of the
population is called a sample. Then we collect the required information from the elements in-
cluded in the sample.
Definition
Sample Survey The technique of collecting information from a portion of the population is
called a sample survey.
A survey can be conducted by personal interviews, by telephone, or by mail. The personal
interview technique has the advantages of a high response rate and a high quality of answers
Even the seemingly simplest of questions can yield complex answers. "Do you own a car?" asks Stanley Presser,
a sociologist at the National Science Foundation in Washington, D.C. "That sounds like an awfully simple question.
But is it really? What does 'you' mean? Suppose a wife is answering the poll, and the car is registered in her hus-
band's name. How is she supposed to answer? What does 'own' mean? What if the car is on a long-term lease?
What does 'car' mean? What if they have one of those new little vans, or a four-wheel-drive vehicle? My God, that
sounds like a simple question! You can imagine how diverse the factors become in a more complicated one."
Suppose, however, that the question about car ownership had been preceded by a series of related ques-
tions: "Are you married? Does your spouse drive an automotive vehicle? Is it a car, a van or some other sort of ve-
hicle? Is it leased, or does your spouse own it? Now about you— do you own a car?" Such a series of questions
would serve to clarify the intended meaning of the one about car ownership.
Source: Rich Jaroslovsky, "What's on Your Mind, America?" Psychology Today, July-August IE
Sussex Publishers, Inc. Reprinted with permission.
8, 54-59. Copyright © 1 i
IS IT A
SIMPLE
QUESTION?
obtained. However, it is the most expensive and time-consuming technique. The telephone sur-
vey also gives a high response rate. It is less expensive and less time-consuming than personal
interviews. Nonetheless, a problem with telephone surveys is that many people do not like to
be called at home, and those who do not have a phone are left out of the survey. A survey con-
ducted by mail is the least expensive method, but the response rate is usually very low. Many
people included in such a survey do not return the questionnaires.
Conducting a survey that gives accurate and reliable results is not an easy task. To quote
Warren Mitofsky, director of Elections and Surveys for CBS News, "Any damn fool with 10
phones and a typewriter thinks he can conduct a poll." 1 Preparing a questionnaire is probably the
most difficult part of a survey. The way a question is phrased can affect the results of the sur-
vey. Case Study A-l, which is excerpted from an article published in Psychology Today, shows
that writing questions for a questionnaire is a much more complex task than is usually thought.
Section A. 2 discusses sample surveys and sampling techniques in detail.
Experiments
In an experiment, we exercise control over some factors when we collect information.
Definition
Experiment In an experiment, data are collected from members of a population or sample in
such a way that we have some control over the factors that may affect the characteristic of in-
terest or the results of the experiment.
For example, how is a new drug to be tested to find out whether or not it cures a disease?
This is done by designing an experiment in which the patients under study are divided into two
groups as follows:
1. The treatment group — the members of this group receive the actual drug.
2. The control group — the members of this group do not receive the actual drug but are given
a substitute (called a placebo) that appears to be the actual drug.
The two groups are formed in such a way that the patients in one group are similar to the
patients in the other group. This is done by making random assignments of patients to the two
groups. Neither the doctors nor the patients know to which group a patient belongs. Such an
experiment is called a double -blind experiment. Then, after a comparison of the percentage of
patients cured in each of the two groups, a decision is made about the effectiveness or noneffec-
tiveness of the new drug. For more on experiments, refer to Section A. 3 on experimental design.
'"The Numbers Racket: How Polls and Statistics Lie," U.S. News & World Report, July 11, 1988.
A4 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
A.2 Sample Surveys and Sampling Techniques
In this section first we discuss the reasons sample surveys are preferred over a census, and then
we discuss a representative sample, random and nonrandom samples, sampling and nonsam-
pling errors, and random sampling techniques.
A.2.1 Why Sample?
As mentioned in the previous section, most of the time surveys are conducted by using samples
and not a census of the population. Three of the main reasons for conducting a sample survey
instead of a census are listed next.
Time
In most cases, the size of the population is quite large. Consequently, conducting a census takes
a long time, whereas a sample survey can be conducted very quickly. It is time-consuming to
interview or contact hundreds of thousands or even millions of members of a population. On
the other hand, a survey of a sample of a few hundred elements may be completed in little time.
In fact, because of the amount of time needed to conduct a census, by the time the census is
completed, the results may be obsolete.
Cost
The cost of collecting information from all members of a population may easily fall outside the
limited budget of most, if not all, surveys. Consequently, to stay within the available resources,
conducting a sample survey may be the best approach.
Impossibility of Conducting a Census
Sometimes it is impossible to conduct a census. First, it may not be possible to identify and ac-
cess each member of the population. For example, if a researcher wants to conduct a survey
about homeless people, it is not possible to locate each member of the population and include
him or her in the survey. Second, sometimes conducting a survey means destroying the items
included in the survey. For example, to estimate the mean life of lightbulbs would necessitate
burning out all the bulbs included in the survey. The same is true about finding the average life
of batteries. In such cases, only a portion of the population can be selected for the survey.
A.2.2 Random and Nonrandom Samples
Depending on how a sample is drawn, it may be a random sample or a nonrandom sample.
Definition
Random and Nonrandom Samples A random sample is a sample drawn in such a way that each
member of the population has some chance of being selected in the sample. In a nonrandom sam-
ple, some members of the population may not have any chance of being selected in the sample.
Suppose we have a list of 100 students and we want to select 10 of them. If we write the
names of all 100 students on pieces of paper, put them in a hat, mix them, and then draw 10 names,
the result will be a random sample of 10 students. However, if we arrange the names of these 100
students alphabetically and pick the first 10 names, it will be a nonrandom sample because the
students who are not among the first 10 have no chance of being selected in the sample.
A random sample is usually a representative sample. Note that for a random sample, each
member of the population may or may not have the same chance of being included in the sam-
ple. Four types of random samples are discussed in Section A. 2.4.
A.2 Sample Surveys and Sampling Techniques A5
Two types of nonrandom samples are a convenience sample and a judgment sample. In a con-
venience sample, the most accessible members of the population are selected to obtain the results
quickly. For example, an opinion poll may be conducted in a few hours by collecting information
from certain shoppers at a single shopping mall. In a judgment sample, the members are selected
from the population based on the judgment and prior knowledge of an expert. Although such a
sample may happen to be a representative sample, the chances of it being so are small. If the pop-
ulation is large, it is not an easy task to select a representative sample based on judgment.
The so-called pseudo polls are examples of nonrepresentative samples. For instance, a survey
conducted by a magazine that includes only its own readers does not usually involve a represen-
tative sample. Similarly, a poll conducted by a television station giving two separate telephone
numbers for yes and no votes is not based on a representative sample. In these two examples, re-
spondents will be only those people who read that magazine or watch that television station, who
do not mind paying the postage or telephone charges, or who feel compelled to respond.
Another kind of sample is the quota sample. To draw such a sample, we divide the target
population into different subpopulations based on certain characteristics. Then we select a sub-
sample from each subpopulation in such a way that each subpopulation is represented in the sam-
ple in exactly the same proportion as in the target population. As an example of a quota sample,
suppose we want to select a sample of 1000 persons from a city whose population has 48% men
and 52% women. To select a quota sample, we choose 480 men from the male population and
520 women from the female population. The sample selected in this way will contain exactly
48% men and 52% women. Another way to select a quota sample is to select from the popula-
tion one person at a time until we have exactly 480 men and 520 women.
Until the 1948 presidential election in the United States, quota sampling was the most com-
monly used sampling procedure to conduct opinion polls. The voters included in the samples were
selected in such a way that they represented the population proportions of voters based on age, sex,
education, income, race, and so on. However, this procedure was abandoned after the 1948 presi-
dential election, in which the underdog, Harry Truman, defeated Thomas E. Dewey, who was heav-
ily favored based on the opinion polls. First, the quota samples failed to be representative because
the interviewers were allowed to fill their quotas by choosing voters based on their own judgments.
This caused the selection of more upper-income and highly educated people, who happened to be
Republicans. Thus, the quota samples were unrepresentative of the population because Republicans
were overrepresented in these samples. Second, the results of the opinion polls based on quota sam-
pling happened to be false because a large number of factors differentiate voters, but the pollsters
considered only a few of those factors. A quota sample based on a few factors will skew the re-
sults. A random sample (one that is not based on quotas) has a much better chance of being rep-
resentative of the population of all voters than a quota sample based on a few factors.
A.2.3 Sampling and Nonsampling Errors
The results obtained from a sample survey may contain two types of errors: sampling and non-
sampling errors. The sampling error is also called the chance error, and nonsampling errors are
also called the systematic errors.
Sampling or Chance Error
Usually, all samples selected from the same population will give different results because they
contain different elements of the population. Moreover, the results obtained from any one sample
will not be exactly the same as the ones obtained from a census. The difference between a sample
result and the result we would have obtained by conducting a census is called the sampling
error, assuming that the sample is random and no nonsampling error has been made.
Definition
Sampling Error The sampling error is the difference between the result obtained from a sam-
ple survey and the result that would have been obtained if the whole population had been in-
cluded in the survey.
A6 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
The sampling error occurs because of chance, and it cannot be avoided. A sampling error
can occur only in a sample survey. It does not occur in a census. Sampling error is discussed
in detail in Section 7.2 of Chapter 7, and an example of it is given there.
Nonsampling or Systematic Errors
Nonsampling errors can occur both in a sample survey and in a census. Such errors occur be-
cause of human mistakes and not chance.
Definition
Nonsampling Errors The errors that occur in the collection, recording, and tabulation of data
are called nonsampling errors.
Nonsampling errors occur because of human mistakes and not chance. Nonsampling errors
can be minimized if questions are prepared carefully and data are handled cautiously. Many types
of systematic errors or biases can occur in a survey, including selection error, nonresponse error,
response error, and voluntary response error. The following chart shows the types of errors.
Types of errors
Sampling or Nonsampling or
chance error systematic errors
Selection Nonresponse Response Voluntary
error error error response
error
(i) Selection Error
When we need to select a sample, we use a list of elements from which we draw a sample,
and this list usually does not include many members of the target population. Most of the
time it is not feasible to include every member of the target population in this list. This list
of members of the population that is used to select a sample is called the sampling frame.
For example, if we use a telephone directory to select a sample, the list of names that appears
in this directory makes the sampling frame. In this case we will miss the people who are not
listed in the telephone directory. The people we miss, for example, will be poor people (in-
cluding homeless people) who do not have telephones and people who do not want to be listed
in the directory. Thus, the sampling frame that is used to select a sample may not be repre-
sentative of the population. This may cause the sample results to be different from the pop-
ulation results. The error that occurs because the sampling frame is not representative of the
population is called the selection error.
Definition
Selection Error The list of members of the target population that is used to select a sample is
called the sampling frame. The error that occurs because the sampling frame is not representa-
tive of the population is called the selection error.
A.2 Sample Surveys and Sampling Techniques A7
If a sample is nonrandom (and, hence, nonrepresentative), the sample results may be quite
different from the census results.
(ii) Nonresponse Error
Even if our sampling frame and, consequently, the sample are representative of the population,
nonresponse error may occur because many of the people included in the sample did not re-
spond to the survey.
Definition
Nonresponse Error The error that occurs because many of the people included in the sample
do not respond to a survey is called the nonresponse error.
This type of error occurs especially when a survey is conducted by mail. A lot of people
do not return the questionnaires. It has been observed that families with low and high incomes
do not respond to surveys by mail. Consequently, such surveys overrepresent middle-income
families. This kind of error occurs in other types of surveys, too. For instance, in a face-to-face
survey where the interviewer interviews people in their homes, many people may not be home
when the interviewer visits their homes. The people who are home at the time the interviewer
visits and the ones who are not home at that time may differ in many respects, causing a bias
in the survey results. This kind of error may also occur in a telephone survey. Many people may
not be home when the interviewer calls. This may distort the results. To avoid the nonresponse
error, every effort should be made to contact all people included in the survey.
(iii) Response Error
The response error occurs when the answer given by a person included in the survey is not
correct. This may happen for many reasons. One reason is that the respondent may not have
understood the question. Thus, the wording of the question may have caused the respondent to
answer incorrectly. It has been observed that when the same question is worded differently,
many people do not respond the same way. Usually such an error on the part of respondents is
not intentional.
Definition
Response Error The response error occurs when people included in the survey do not provide
correct answers.
Sometimes the respondents do not want to give correct information when answering a ques-
tion. For example, many respondents will not disclose their true incomes on questionnaires or
in interviews. When information on income is provided, it is almost always biased in the up-
ward direction.
Sometimes the race of the interviewer may affect the answers of respondents. This is es-
pecially true if the questions asked are about race relations. The answers given by respondents
may differ depending on the race of the interviewer.
(iv) Voluntary Response Error
Another source of systematic error is a survey based on a voluntary response sample.
Definition
Voluntary Response Error Voluntary response error occurs when a survey is not conducted on
a randomly selected sample but a questionnaire is published in a magazine or newspaper and
people are invited to respond to that questionnaire.
A8 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
The polls conducted based on samples of readers of magazines and newspapers suffer from
voluntary response error or bias. Usually only those readers who have very strong opinions about
the issues involved respond to such surveys. Surveys in which the respondents are required to call
some telephone numbers also suffer from this type of error. Here, to participate, many times a
respondent have to pay for the call, and many people do not want to bear this cost. Consequently,
the sample is usually neither random nor representative of the target population because partici-
pation is voluntary.
A.2.4 Random Sampling Techniques
There are many ways to select a random sample. Four of these techniques are discussed next.
Simple Random Sampling
Under this sampling technique, each sample of the same size selected from the same popula-
tion has the same probability of being selected.
Definition
Simple Random Sampling In this sampling technique, each sample of the same size has the
same probability of being selected. Such a sample is called a simple random sample.
One way to select a simple random sample is by a lottery or drawing. For example, if we need
to select 5 students from a class of 50, we write each of the 50 names on a separate piece of
paper. Then, we place all 50 names in a hat and mix them thoroughly. Next, we draw 1 name
randomly from the hat. We repeat this experiment four more times. The 5 drawn names make
up a simple random sample.
The second procedure to select a simple random sample is to use a table of random num-
bers, which has become an outdated procedure. In this age of technology, it is much easier to
use a statistical package, such as Minitab, to select a simple random sample.
Systematic Random Sampling
The simple random sampling procedure becomes very tedious if the size of the population is
large. For example, if we need to select 150 households from a list of 45,000, it is very time-
consuming either to write the 45,000 names on pieces of paper and then select 150 households
or to use a table of random numbers. In such cases, it is more convenient to use systematic
random sampling.
The procedure to select a systematic random sample is as follows. In the example just men-
tioned, we would arrange all 45,000 households alphabetically (or based on some other char-
acteristic). Since the sample size should equal 150, the ratio of population to sample size is
45,000/150 = 300. Using this ratio, we randomly select one household from the first 300 house-
holds in the arranged list using either method. Suppose by using either of the methods, we
select the 210th household. We then select every 210th household from every 300 households
in the list. In other words, our sample includes the households with numbers 210, 510, 810,
1110, 1410, 1710, and soon.
Definition
Systematic Random Sample In systematic random sampling, we first randomly select one mem-
ber from the first k units. Then every kth member, starting with the first selected member, is in-
cluded in the sample.
Stratified Random Sampling
Suppose we need to select a sample from the population of a city, and we want households with
different income levels to be proportionately represented in the sample. In this case, instead of
A3 Design of Experiments A9
selecting a simple random sample or a systematic random sample, we may prefer to apply a dif-
ferent technique. First, we divide the whole population into different groups based on income
levels. For example, we may form three groups of low-, medium-, and high-income households. We
will now have three subpopulations, which are usually called strata. We then select one sample
from each subpopulation or stratum. The collection of all three samples selected from three strata
gives the required sample, called the stratified random sample. Usually, the sizes of the samples
selected from different strata are proportionate to the sizes of the subpopulations in these strata. Note
that the elements of each stratum are identical with regard to the possession of a characteristic.
Definition
Stratified Random Sample In a stratified random sample, we first divide the population into sub-
populations, which are called strata. Then, one sample is selected from each of these strata. The
collection of all samples from all strata gives the stratified random sample.
Thus, whenever we observe that a population differs widely in the possession of a charac-
teristic, we may prefer to divide it into different strata and then select one sample from each
stratum. We can divide the population on the basis of any characteristic, such as income, ex-
penditure, sex, education, race, employment, or family size.
Cluster Sampling
Sometimes the target population is scattered over a wide geographical area. Consequently, if a
simple random sample is selected, it may be costly to contact each member of the sample. In
such a case, we divide the population into different geographical groups or clusters and as a
first step select a random sample of certain clusters from all clusters. We then take a random
sample of certain elements from each selected cluster. For example, suppose we are to conduct
a survey of households in the state of New York. First, we divide the whole state of New York
into, say, 40 regions, which are called clusters or primary units. We make sure that all clus-
ters are similar and, hence, representative of the population. We then select at random, say,
5 clusters from 40. Next, we randomly select certain households from each of these 5 clusters
and conduct a survey of these selected households. This is called cluster sampling. Note that
all clusters must be representative of the population.
Definition
Cluster Sampling In cluster sampling, the whole population is first divided into (geographical)
groups called clusters. Each cluster is representative of the population. Then a random sample
of clusters is selected. Finally, a random sample of elements from each of the selected clusters
is selected.
A.3 Design of Experiments
As mentioned earlier, to use statistical methods to make decisions, we need access to data. Con-
sider the following examples about decision making.
1. A government agency wants to find the average income of households in the United States.
2. A company wants to find the percentage of defective items produced on a machine.
3. A researcher wants to know if there is an association between eating unhealthy food and
cholesterol level.
4. A pharmaceutical company has developed a new medicine for a disease and it wants to
check if this medicine cures the disease.
All of these cases relate to decision making. We cannot reach a conclusion in these examples
unless we have access to data. Data can be obtained from observational studies, experiments, or
A10 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
surveys. This section is devoted mainly to controlled experiments. However, it also explains ob-
servational studies and how they differ from surveys.
Suppose two diets, Diet 1 and Diet 2, are being promoted by two different companies, and
each of these companies claims that its diet is successful in reducing weight. A research nutri-
tionist wants to compare these diets with regard to their effectiveness for losing weight.
Following are the two alternatives for the researcher to conduct this research.
1. The researcher contacts the persons who are using these diets and collects information on
their weight loss. The researcher may contact as many persons as she has the time and fi-
nancial resources for. Based on this information, the researcher makes a decision about the
comparative effectiveness of these diets.
2. The researcher selects a sample of persons who want to lose weight, divides them randomly
into two groups, and assigns each group to one of the two diets. Then she compares these
two groups with regard to the effectiveness of these diets.
The first alternative is an example of an observational study, and the second is an example
of a controlled experiment.
Definition
Treatment A condition (or a set of conditions) that is imposed on a group of elements by the
experimenter is called a treatment.
In an observational study the investigator does not impose a treatment on subjects or ele-
ments included in the study. For instance, in the first alternative, the researcher simply
collects information from the persons who are currently using these diets. In this case, the
persons were not assigned to the two diets at random; instead, they chose the diets voluntarily.
In this situation the researcher's conclusion about the comparative effectiveness of the two
diets may not be valid because the effects of the diets will be confounded with many other
factors or variables. When the effects of one factor cannot be separated from the effects of
some other factors, the effects are said to be confounded. The persons who chose Diet 1 may
be completely different with regard to age, gender, and eating and exercise habits from the
persons who chose Diet 2. Thus, the weight loss may not be due entirely to the diet but to
other factors or variables as well. Persons in one group may aggressively manage both diet
and exercise, for example, whereas persons in the second group may depend entirely on diet.
Thus, the effects of these other variables will get mixed up (confounded) with the effect of
the diets.
Under the second alternative, the researcher selects a group of people, say 100, and ran-
domly assigns them to two diets. One way to make random assignments is to write the name
of each of these persons on a piece of paper, put them in a hat, and then randomly draw 50
names from this hat. These 50 persons will be assigned to one of the two diets, say Diet 1. The
remaining 50 persons will be assigned to the second diet, Diet 2. This procedure is called
randomization. Note that random assignments can also be made by using other methods such
as a table of random numbers or technology.
Definition
Randomization The procedure in which elements are assigned to different groups at random
is called randomization.
When people are assigned to one or the other of two diets at random, the other differences
among people in the two groups almost disappear. In this case these groups will not differ very
much with regard to such factors as age, gender, and eating and exercise habits. The two groups
A.3 Design of Experiments All
will be very similar to each other. By using the random process to assign people to one or the
other of two diets, we have controlled the other factors that can affect the weights of people.
Consequently, this is an example of a designed experiment.
As mentioned earlier, a condition (or a set of conditions) that is imposed on a group of ele-
ments by the experimenter is called a treatment. In the example on diets, each of the two diet
types is called a treatment. The experimenter randomly assigns the elements to these two treat-
ments. Again, in such cases the study is called a designed experiment.
Definition
Designed Experiment and Observational Study When the experimenter controls the (random)
assignment of elements to different treatment groups, the study is said to be a designed exper-
iment. In contrast, in an observational study the assignment of elements to different treatments
is voluntary, and the experimenter simply observes the results of the study.
The group of people who receive a treatment is called the treatment group, and the group
of people who do not receive a treatment is called the control group. In our example on diets,
both groups are treatment groups because each group is assigned to one of the two types of diet.
That example does not contain a control group.
Definition
Treatment and Control Groups The group of elements that receives a treatment is called the
treatment group, and the group of elements that does not receive a treatment is called the con-
trol group.
■ EXAMPLE A-l
Suppose a pharmaceutical company has developed a new medicine to cure a disease. To see
whether or not this medicine is effective in curing this disease, it will have to be tested on a
group of humans. Suppose there are 100 persons who have this disease; 50 of them voluntarily
decide to take this medicine, and the remaining 50 decide not to take it. The researcher then
compares the cure rates for the two groups of patients. Is this an example of a designed ex-
periment or an observational study?
Solution This is an example of an observational study because 50 patients voluntarily joined
the treatment group; they were not randomly selected. In this case, the results of the study may
not be valid because the effects of the medicine will be confounded with other variables. All
of the patients who decided to take the medicine may not be similar to the ones who decided
not to take it. It is possible that the persons who decided to take the medicine are in the ad-
vanced stages of the disease. Consequently, they do not have much to lose by being in the
treatment group. The patients in the two groups may also differ with regard to other factors
such as age, gender, and so on. I
An example of an
observational study.
■ EXAMPLE A-2
Reconsider Example A-l. Now, suppose that out of the 100 people who have this disease, 50
are selected at random. These 50 people make up one group, and the remaining 50 belong to
the second group. One of these groups is the treatment group, and the second is the control
An example of a
designed experiment.
A12 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
group. The researcher then compares the cure rates for the two groups of patients. Is this an
example of a designed experiment or an observational study?
Solution In this case, the two groups will be very similar to each other. Note that we do
not expect the two groups to be exactly identical. However, when randomization is used,
the two groups will be very similar. After these two groups have been formed, one group
will be given the actual medicine. This group is called the treatment group. The other group
will be administered a placebo (a dummy medicine that looks exactly like the actual med-
icine). This group is called the control group. This is an example of a designed experiment
because the patients are assigned to one of two groups — the treatment or the control group —
randomly. I
Usually in an experiment like the one in Example A-2, patients do not know which group
they belong to. Most of the time the experimenters do not know which group a patient
belongs to. This is done to avoid any bias or distortion in the results of the experiment. When
neither patients nor experimenters know who is taking the real medicine and who is taking
the placebo, it is called a double-blind experiment. For the results of the study to be unbi-
ased and valid, an experiment must be a double-blind designed experiment. Note that if either
experimenters or patients or both have access to information regarding which patients belong
to treatment or control groups, it will no longer be a double -blind experiment.
The use of placebos in medical experiments is very important. A placebo is just a dummy
pill that looks exactly like the real medicine. Often, patients respond to any kind of medicine.
Many studies have shown that even when the patients were given sugar pills (and did not
know it), many of them indicated a decrease in pain. Patients respond to placebos because
they have confidence in their physicians and medicines. This is called the placebo effect.
Note that there can be more than two groups of elements in an experiment. For example, an
investigator may need to compare three diets for chickens with regard to weight gain. Here, in a
designed experiment, the chickens will be randomly assigned to one of the three diets, which are
the three treatments.
In some instances we have to base our research on observational studies because it is not
feasible to conduct a designed experiment. For example, suppose a researcher wants to compare
the starting salaries of business and psychology majors. The researcher will have to depend on
an observational study. She will select two samples, one of recent business majors and another
of recent psychology majors. Based on the starting salaries of these two groups, the researcher
will make a decision. Note that, here, the effects of the majors on the starting salaries of the two
groups of graduates will be confounded with other variables. One of these other factors is that
the business and psychology majors may be different in regard to intelligence level, which may
affect their salaries. However, the researcher cannot conduct a designed experiment in this case.
She cannot select a group of persons randomly and ask them to major in business and select
another group and ask them to major in psychology. Instead, persons voluntarily choose their
majors.
In a survey we do not exercise any control over the factors when we collect information.
This characteristic of a survey makes it very close to an observational study. However, a sur-
vey may be based on a probability sample, which differentiates it from an observational
study.
If an observational study or a survey indicates that two variables are related, it does not mean
that there is a cause-and-effect relationship between them. For example, if an economist takes a
sample of families, collects data on the incomes and rents paid by these families, and establishes
an association between these two variables, it does not necessarily mean that families with high-
er incomes pay higher rents. Here the effects of many variables on rents are confounded. A fam-
ily may pay a higher rent not because of higher income but because of various other factors, such
as family size, preferences, or place of residence. We cannot make a statement about the cause-
and-effect relationship between incomes and rents paid by families unless we control for these
other variables. The association between incomes and rents paid by families may fit any of the
following scenarios.
Antibacterial soaps are no better than regular soap. Experts have said so for years. But that has not stopped
millions of Americans from snapping up the supposedly superior germ killers— now 76 percent of the liquid-
soap market. Part of the problem was the lack of rigorous studies to back up the experts' claims. But last week
[end of October 2002] at the annual meeting of the Infectious Diseases Society of America, Elaine Larson, as-
sociate dean for research at Columbia University's School of Nursing, came up with the goods. In a random-
ized, double-blind, controlled study— the type of trial used to test pharmaceuticals— she surveyed 224 New York
City homemakers. Half were given ordinary liquid soaps for a full year and the other half received antibacterial
soaps. All participants' hands were cultured for germs at the beginning and the end of the study.
The results? At the outset, all participants' hands were teeming with 800,000 to 1 million bacteria. "That's
normal," says Larson. "People can have up to 10 million on their hands." By the end of the year, tests revealed
that they had just 300,000 or so. It didn't matter whether they used antibacterial soap or not. The difference
was that they were taking more time to wash their hands thoroughly, particularly the fingers, which come in con-
tact with the most foreign objects during the day.
Why don't antibacterial soaps do better? "The antimicrobial agent triclosan requires several minutes of con-
tact to work," says Dr. Stuart Levy of Tufts University, author of "The Antibiotic Paradox." "Most people wash their
hands for three to five seconds." Unfortunately, residues of antimicrobial soaps do linger on sinks and countertops,
where Levy says they may contribute to the development of drug-resistant bacteria. A better solution for people
with babies or immune-compromised patients at home is to use an alcohol-based gel, which kills germs by dry-
ing them out. Last week [end of October 2002] the CDC recommended these waterless germicides even in hos-
pitals. Now, that's what the doctor ordered.
DO ANTI-
BACTERIAL
SOAPS
WORK?
Source: Anne Underwood, "The Real
Dirt on Antibacterial Soaps." Newsweek,
November 4, 2002. Reproduced with
permission.
These two variables have a cause-and-effect relationship. Families that have higher in-
comes do pay higher rents. A change in incomes of families causes a change in rents paid.
The incomes and rents paid by families do not have a cause-and-effect relationship. Both
of these variables have a cause-and-effect relationship with a third variable. Whenever that
third variable changes, these two variables change.
The effect of income on rent is confounded with other variables, and this indicates that in-
come affects rent paid by families.
If our purpose in a study is to establish a cause-and-effect relationship between two variables,
we must control for the effects of other variables. In other words, we must conduct a designed study.
EXERCISES
A.l Briefly describe the various sources of data.
A.2 What is the difference between internal and external sources of data? Explain.
A.3 Explain the difference between a sample survey and a census. Why is a sample survey usually pre-
ferred over a census?
A.4 What is the difference between a survey and an experiment? Explain.
A.5 Explain the following.
a. Random sample b. Nonrandom sample c. Convenience sample
d. Judgment sample e. Quota sample
A.6 Explain briefly the following four sampling techniques.
a. Simple random sampling b. Systematic random sampling
c. Stratified random sampling d. Cluster sampling
A.7 In which sampling technique do all samples of the same size selected from a population have the
same chance of being selected?
A.8 A statistics professor wanted to find out the average GPA (grade point average) for all students at her
university. She used all students enrolled in her statistics class as a sample and collected information on
their GPAs to find the average GPA.
a. Is this sample a random or a nonrandom sample? Explain.
b. What kind of sample is it? In other words, is it a simple random sample, a systematic sample, a strat-
ified sample, a cluster sample, a convenience sample, a judgment sample, or a quota sample? Explain.
c. What kind of systematic error, if any, will be made with this kind of sample? Explain.
An
AT 4 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
A.9 A professor wanted to select 20 students from his class of 300 students to collect detailed informa-
tion on the profiles of his students. He used his knowledge and expertise to select these 20 students.
a. Is this sample a random or a nonrandom sample? Explain.
b. What kind of sample is it? In other words, is it a simple random sample, a systematic sample, a
stratified sample, a cluster sample, a convenience sample, a judgment sample, or a quota sample?
Explain.
c. What kind of systematic error, if any, will be made with this kind of sample? Explain.
A.10 Refer to Exercise A.8. Suppose the professor obtains a list of all students enrolled at the university
from the registrar's office and then selects 150 students at random from this list using a statistical soft-
ware package such as Minitab.
a. Is this sample a random or a nonrandom sample? Explain.
b. What kind of sample is it? In other words, is it a simple random sample, a systematic sample, a
stratified sample, a cluster sample, a convenience sample, a judgment sample, or a quota sam-
ple? Explain.
c. Do you think any systematic error will be made in this case? Explain.
A.ll Refer to Exercise A.9. Suppose the professor enters the names of all students enrolled in his class
on a computer. He then selects a sample of 20 students at random using a statistical software package such
as Minitab.
a. Is this sample a random or a nonrandom sample? Explain.
b. What kind of sample is it? In other words, is it a simple random sample, a systematic sample, a
stratified sample, a cluster sample, a convenience sample, a judgment sample, or a quota sam-
ple? Explain.
c. Do you think any systematic error will be made in this case? Explain.
A.12 A company has 1000 employees, of whom 58% are men and 42% are women. The research de-
partment at the company wanted to conduct a quick survey by selecting a sample of 50 employees and
asking them about their opinions on an issue. They divided the population of employees into two groups,
men and women, and then selected 29 men and 21 women from these respective groups. The interviewers
were free to choose any 29 men and 21 women they wanted. What kind of sample is it? Explain.
A.13 A magazine published a questionnaire for its readers to fill out and mail to the magazine's office.
In the questionnaire, cell phone owners were asked how much they would have to be paid to do without
their cell phones for one month. The magazine received responses from 5439 cell phone owners.
a. Based on the discussion of types of samples in Section A. 2. 2, what type of sample is this? Ex-
plain.
b. To what kind(s) of systematic error, if any, would this survey be subject?
A.14 A researcher wanted to conduct a survey of major companies to find out what benefits are offered
to their employees. She mailed questionnaires to 2500 companies and received questionnaires back from
493 companies. What kind of systematic error does this survey suffer from? Explain.
A. 15 An opinion poll agency conducted a survey based on a random sample in which the interviewers
called the parents included in the sample and asked them the following questions:
i. Do you believe in spanking children?
ii. Have you ever spanked your children?
iii. If the answer to the second question is yes, how often?
What kind of systematic error, if any, does this survey suffer from? Explain.
A.16 A survey based on a random sample taken from a borough of New York City showed that 65% of
the people living there would prefer to live somewhere other than New York City if they had the oppor-
tunity to do so. Based on this result, can the researcher say that 65% of people living in New York City
would prefer to live somewhere else if they had the opportunity to do so? Explain.
A.17 In March 2005, the New England Journal of Medicine published the results of a 10-year clinical
trial of low-dose aspirin therapy for the cardiovascular health of women {Time, March 21, 2005). The study
was based on 40,000 healthy women, most of whom were in their 40s and 50s when the trial began. Half
of these women were administered 100 mg of aspirin every other day, and the others were given a placebo.
Assume that the women were assigned randomly to these two groups.
a. Is this an observational study or a designed experiment? Explain.
b. From the information given above, can you determine whether or not this is a double-blind study?
Explain. If not, what additional information would you need?
A. 18 Refer to Exercise A.17. That study also looked at the incidences of heart attacks in the two groups
of women. Overall the study did not find a statistically significant difference in heart attacks between the
two groups of women. However, the study noted that among women who were at least 65 years old when
A.3 Design of Experiments A15
the study began, there was a lower incidence of heart attack for those who took aspirin than for those who
took a placebo. Suppose that some medical researchers want to study this phenomenon more closely. They
recruit 2000 healthy women aged 65 years and older, and randomly divide them into two groups. One group
takes 100 mg of aspirin every other day, and the other group takes a placebo. The women did not know to
which group they belonged, but the doctors who conducted the study had access to this information.
a. In this an observational study or a designed experiment? Explain.
b. Is this a double-blind study? Explain.
A.19 Refer to Exercise A. 18. Now suppose that neither patients nor doctors knew what group patients be-
longed to.
a. Is this an observational study or a designed experiment? Explain.
b. Is this study a double-blind study? Explain.
A.20 A federal government think tank wanted to investigate whether a job training program helps the
families who are on welfare to get off the welfare program. The researchers at this agency selected 5000
volunteer families who were on welfare and offered the adults in those families free job training. The
researchers selected another group of 5000 volunteer families who were on welfare and did not offer
them such job training. After 3 years the two groups were compared in regard to the percentage of fam-
ilies who got off welfare. Is this an observational study or a designed experiment? Explain.
A.21 Refer to Exercise A.20. Now suppose the agency selected 10,000 families at random from the list
of all families that were on welfare. Of these 10,000 families, the agency randomly selected 5000 fami-
lies and offered them free job training. The remaining 5000 families were not offered such job training.
After 3 years the two groups were compared in regard to the percentage of families who got off welfare.
Is this an observational study or a designed experiment? Explain.
A.22 Refer to Exercise A.20. Based on that study, the researchers concluded that the job training program
causes (helps) families who are on welfare to get off the welfare program. Do you agree with this con-
clusion? Explain.
A.23 Refer to Exercise A.21. Based on that study, the researchers concluded that the job training program
causes (helps) families who are on welfare to get off the welfare program. Do you agree with this con-
clusion? Explain.
A.24 A researcher advertised for volunteers to study the relationship between the amount of meat consumed
and cholesterol level. In response to this advertisement, 3476 persons volunteered. The researcher collected
information on the meat consumption and cholesterol level of each of these persons. Based on these data,
the researcher concluded that there is a very strong positive association between these two variables.
a. Is this an observational study or a designed experiment? Explain.
b. Based on this study, can the researcher conclude that consumption of meat increases cholesterol
level? Explain why or why not.
A.25 A pharmaceutical company developed a new medicine for compulsive behavior. To test this medi-
cine on humans, the company advertised for volunteers who were suffering from this disease and wanted
to participate in the study. As a result, 1820 persons responded. Using their own judgment, the group of
physicians who were conducting this study assigned 910 of these patients to the treatment group and the
remaining 910 to the control group. The patients in the treatment group were administered the actual med-
icine, and the patients in the control group were given a placebo. Six months later the conditions of the
patients in the two groups were examined and compared. Based on this comparison, the physicians con-
cluded that this medicine improves the condition of patients suffering from compulsive behavior.
a. Comment on this study and its conclusion.
b. Is this an observational study or a designed experiment? Explain.
c. Is this a double-blind study? Explain.
A.26 Refer to Exercise A.25. Suppose the physicians conducting this study obtained a list of all patients
suffering from compulsive behavior who were being treated by doctors in all hospitals in the country. Fur-
ther assume that this list is representative of the population of all such patients. The physicians then ran-
domly selected 1820 patients from this list. Of these 1820, a randomly selected group of 910 patients were
assigned to the treatment group, and the remaining 910 patients were assigned to the control group. The
patients did not know which group they belonged to, but the doctors had access to such information. Six
months later the conditions of the patients in the two groups were examined and compared. Based on this
comparison, the physicians concluded that this medicine improves the condition of patients suffering from
compulsive behavior.
a. Comment on this study and its conclusion.
b. Is this an observational study or a designed experiment? Explain.
c. Is this a double-blind study? Explain.
A16 Appendix A Sample Surveys, Sampling Techniques, and Design of Experiments
A.27 Refer to Exercise A.26. Now suppose that neither patients nor doctors knew what group the patients
belonged to.
a. Is this an observational study or a designed experiment? Explain.
b. Is this a double-blind study? Explain.
A.28 The Centre for Nutrition and Food Research at Queen Margaret University College in Edinburgh
studied the relationship between sugar consumption and weight gain {Fitness, May 2002). All the people
who participated in the study were divided into two groups, and both of these groups were put on low-
calorie, low-fat diets. The diet of the people in the first group was low in sugar, but the people in the
second group received as much as 10% of their calories from sucrose. Both groups stayed on their
respective diets for 8 weeks. During these 8 weeks, participants in both groups lost 1/2 to 3/4 pound
per week.
a. Was this a designed experiment or an observational study?
b. Was there a control group in this study?
c. Was this a double-blind experiment?
A.29 A psychologist needs 10 pigs for a study of the intelligence of pigs. She goes to a pig farm where
there are 40 young pigs in a large pen. Assume that these pigs are representative of the population of all
pigs. She selects the first 10 pigs she can catch and uses them for her study.
a. Do these 10 pigs make a random sample?
b. Are these 10 pigs likely to be representative of the entire population? Why or why not?
c. If these 10 pigs do not form a random sample, what type of sample is it?
d. Can you suggest a better procedure for selecting a sample of 10 from the 40 pigs in the pen?
A.30 A newspaper wants to conduct a poll to estimate the percentage of its readers who favor a gambling
casino in their city. People register their opinions by placing a phone call that costs them $ 1 .
a. Is this method likely to produce a random sample?
b. Which, if any, of the types of biases listed in this appendix are likely to be present and why?
■ ADVANCED EXERCISES
A.31 A researcher sent out questionnaires to 5000 randomly chosen members of HMOs (health main-
tenance organizations). Only 1200 of these members completed their questionnaires and returned them.
Seventy-eight percent of the respondents reported that they had experienced denial of claims by their
HMOs. Of those who experienced such denials, 25% had been unable to resolve the problem to their
satisfaction in at least one such instance. Write an article for a business magazine summarizing the
results of the survey and cautioning the readers about possible bias in the results. Indicate which types
of biases are likely to be present, how they could arise, and whether the percentages given above are
likely to overestimate the true percentages of all HMO members who have experienced the denial of
claims by HMOs.
A.32 A college is planning to finance an expansion of its student center through a special $20 annual fee
to be levied on each student for the next 4 years. Because the project will take 2 years to complete, the
students who are currently juniors or seniors will not benefit from the expansion. The campus newspaper
wants to conduct a poll to seek the opinions of students on this expansion. Such opinions of students are
likely to depend on their current class status, so the newspaper decides to use a stratified random sample
with four class levels (freshmen, sophomores, juniors, seniors) as strata. The current student body consists
of 4000 freshmen, 3200 sophomores, 2800 juniors, and 2000 seniors. The sample will contain a total of
300 students, and the size of the sample from each stratum is to be proportional to the size of the sub-
population in each stratum.
a. How many freshmen should be in the sample?
b. How many students should be chosen from each of the other three class levels?
A.33 A college mailed a questionnaire to all 5432 of its alumni who graduated in the last 5 years. One
of the questions was about the current annual incomes of these alumni. Only 1620 of these alumni re-
turned the completed questionnaires, and 1240 of them answered that question. The current mean annual
income of these 1240 respondents was $61,200.
a. Do you think $61,200 is likely to be an unbiased estimate of the current mean annual income of
all 5432 alumni? If so, explain why.
b. If you think that $61,200 is probably a biased estimate of the current mean annual income of all
5432 alumni, what sources of systematic errors discussed in Section A. 2. 3 do you think are pres-
ent here?
c. Do you expect the estimate of $61,200 to be above or below the current mean annual income of
all 5432 alumni? Explain.
Glossary A17
A.34 A group of veterinarians wants to test a new canine vaccine for Lyme disease. (Lyme disease is
transmitted by the bite of an infected deer tick.) One hundred dogs are randomly selected to receive the
vaccine (with their owners' permission) from an area that has a high incidence of Lyme disease. These
dogs are examined by veterinarians for symptoms of Lyme disease once a month for a period of 12 months.
During this 12-month period, 10 of these 100 dogs are diagnosed with Lyme disease. During the same
12-month period, 18% of the unvaccinated dogs in the area are found to have contracted Lyme disease.
a. Does this experiment have a control group?
b. Is this a double-blind experiment?
c. Identify any potential sources of bias in this experiment.
d. Explain how this experiment could have been designed to reduce or eliminate the bias pointed
out in part c.
Glossary
Census A survey conducted by including every element of the pop-
ulation.
Cluster A subgroup (usually geographical) of the population that
is representative of the population.
Cluster sampling A sampling technique in which the population
is divided into clusters and a sample is chosen from one or a few
clusters.
Control group The group on which no condition is imposed.
Convenience sample A sample that includes the most accessible
members of the population.
Designed experiment A study in which the experimenter controls
the assignment of elements to different treatment groups.
Double-blind experiment An experiment in which neither the
doctors (or researchers) nor the patients (or members) know to which
group a patient (or member) belongs.
Experiment A method of collecting data by controlling some or
all factors.
Judgment sample A sample that includes the elements of the pop-
ulation selected based on the judgment and prior knowledge of an
expert.
Nonresponse error The error that occurs because many of the peo-
ple included in the sample do not respond.
Nonsampling or systematic errors The errors that occur in the
collection, recording, and tabulation of data.
Observational study A study in which the assignment of elements
to different treatments is voluntary, and the researcher simply ob-
serves the results of the study.
Quota sample A sample selected in such a way that each group
or subpopulation is represented in the sample in exactly the same
proportion as in the target population.
Random sample A sample that assigns some chance of being se-
lected in the sample to each member of the population.
Randomization The procedure in which elements are assigned to
different (treatment and control) groups at random.
Representative sample A sample that contains the characteristics
of the population as closely as possible.
Response error The error that occurs because people included in
the survey do not provide correct answers.
Sample A portion of the population of interest.
Sample survey A survey that includes elements of a sample.
Sampling frame The list of elements of the target population that
is used to select a sample.
Sampling or chance error The difference between the result ob-
tained from a sample survey and the result that would be obtained
from the census.
Selection error The error that occurs because the sampling frame
is not representative of the population.
Simple random sampling If all samples of the same size selected
from a population have the same chance of being selected, it is called
simple random sampling. Such a sample is called a simple random
sample.
Stratified random sampling A sampling technique in which the
population is divided into different strata and a sample is chosen
from each stratum.
Stratum A subgroup of the population whose members are iden-
tical with regard to the possession of a characteristic.
Survey Collecting data from the elements of a population or
sample.
Systematic random sampling A sampling method used to choose
a sample by selecting every kth unit from the list.
Target population The collection of all subjects of interest.
Treatment A condition (or a set of conditions) that is imposed on
a group of elements by the experimenter. This group is called the
treatment group.
Voluntary response error The error that occurs because a survey
is not conducted on a randomly selected sample, but people are in-
vited to respond voluntarily to the survey.
Appendix
Explanation of Data Sets
This textbook is accompanied by eight large data sets that can be used for statistical analysis
using technology. These data sets are:
Data Set I City Data
Data Set II Data on States
Data Set III NBA Data
Data Set IV Population Data on Manchester Road Race
Data Set V Sample of 500 Observations Selected From Manchester Road Race Data
Data Set VI Data on Movies
Data Set VII Standard & Poor's 100 Index Data
Data Set VIII McDonald's Data
These data sets are available in Minitab, Excel, and Text format on the Web site for this text,
www.wiley.com/college/mann. These data sets can be downloaded from this Web site. If you
need more information on these data sets, you may either contact John Wiley's area represen-
tative or send an email to the author (see Preface). The Web site contains the following files:
1. CITYDATA (This file contains Data Set I)
2. STATEDATA (This file contains Data Set II)
3. NBA (This file contains Data Set III)
4. ROADRACE (This file contains the population data for Data Set IV)
5. RRSAMPLE (This file contains Data Set V)
6. MOVIEDATA (This file contains Data Set VI)
7. S&PDATA (This file contains Data Set VII)
8. MCDONALDDATA (This file contains Data Set VIII)
The extensions MTW, XLS, and TXT indicate that the files are in Minitab, Excel, and Text
formats, respectively.
The following are the explanations of these data sets.
Data Set I: City Data
This data set contains prices (in dollars) of selected products for selected cities across the
country. This data set is reproduced from the ACCRA Cost of Living Index Survey for the
second quarter 2009. It is reproduced with the permission of American Chamber of Commerce
Researchers Association. This data set has 25 columns that contain the following variables.
CI Name of the city
C2 Price of T-bone steak per pound
Bl
B2 Appendix B Explanation of Data Sets
C3 Price of sausage per pound, Jimmy Dean or Owens brand, 100% pork
C4 Price of half-gallon carton of whole milk
C5 Price of parmesan cheese, grated 8 oz. canister, Kraft brand
C6 Price of potatoes, 10 pounds, white or red
C7 Price of fresh orange juice, 64 oz., Tropicana or Florida Natural brand
C8 Price of 75 oz. Cascade dishwashing powder
C9 Price of 16 oz. whole-kernel frozen corn, lowest price
CIO Price of 2 liter Coca Cola, excluding any deposit
Cll Monthly rent of an unfurnished two-bedroom apartment (excluding all utilities ex-
cept water), VA or 2 baths, 950 square feet
C12 Purchase price of 2400 square feet living area new house, on 8000 square feet lot in
urban area with all utilities
C13 Monthly telephone charges for a private residential line; customer owns instruments.
Price includes basic monthly rate; additional local use charges, if any, incurred by a
family of four; TouchTone fee; all other mandatory monthly charges, such as long-
distance access fee and 911 fee; and all taxes on the foregoing.
C14 Price of 1 gallon regular unleaded gas, national brand, including all taxes; cash price
at self-service pump if available
C15 Price of a woman's shampoo, trim, and blow-dry
C16 Price of dry cleaning, man's two-piece suit
C17 Price of first-run movie, indoor, evening, no discount
C18 Price of 1.5 liter bottle of wine, Livingston Cellars or Gallo Chablis or Chenin
Blanc.
C19 Cost of an 11.5 ounce can or brick of coffee
C20 Cost of a one pound box of granulated sugar
C21 Mortgage rate for a conventional 30 year mortgage
C22 Cost of a visit to the doctor's office for a routine examination for a problem with
low to moderate severity
C23 Cost of a tooth-cleaning visit to the dentist's office (established patients only)
C24 Cost of Advil, 200 mg tablets, 100 count
C25 Cost of an 11 or 12 inch thin-crust regular cheese pizza (no extra cheese) at Pizza
Hut and/or Pizza Inn
Data Set II: Data on States
This data set contains information on different variables for all 50 states of the United States
and the district of Columbia. This data set has seven columns that contain the following
variables:
CI Name of the state
C2 Per capita personal income (in current dollars), 2008 (Source: U.S. Bureau of
Economic Analysis)
C3 Traffic fatalities, 2008 (Source: U.S. National Highway Safety Traffic Administration)
C4 Labor force participation rate (in percent), July 2009 (Source: U.S. Bureau of Labor
Statistics)
C5 Average salaries of teachers (in dollars), 2007-08 (Source: Current NEA Estimates
Data Base)
C6 Percent of the population (25 years and older) with a bachelor's degree or higher,
2004-07 (U.S. Census Bureau)
C7 Location (East/West of the Mississippi River)
Data Set VI: Data on Movies B3
Data Set III: NBA Data
This data set contains information on players who were on the rosters of National Basketball
Association (NBA) teams as of April 17, 2009. This data set has 13 columns that contain the
following variables:
CI Team name
C2 Name of player
C3 Player's annual salary
C4 Length of player's contract (in years)
C5 Total value of contract over the length specified in C4
C6 Year contract expires
C7 Number of years of experience in the NBA at the end of the 2008-2009 regular
season
C8 Primary position (C = Center, F = Forward, G = Guard)
C9 Secondary position (blank implies no secondary position)
C10 Height (in inches) of player
Cll Weight (in pounds) of player
C12 Age of player as of April 17, 2009
C13 Identifies whether player came to NBA from college, a foreign country, or high
school
Data Set IV: Manchester (Connecticut)
Road Race Data
This data set contains information on the people who completed the 72nd Annual Thanksgiv-
ing Day Road Race held on November 27, 2008 in Manchester, Connecticut. The total dis-
tance of this race is 4.748 miles, and it is held every year on Thanksgiving Day. A total of
10,431 individuals completed the race. The data set contains three columns that contain the
following variables:
CI Gender (M/F)
C2 Age (in years)
C3 Time to complete the race (in seconds)
Data Set V: Sample of 500 Observations Selected
From Data Set IV
This data set contains a random sample of 500 observations selected from Data Set IV. It has
three columns containing the same variables as listed in Data Set IV.
Data Set VI: Data on Movies
This data set contains information on the top 150 films from 2008 in terms of gross revenue in
the United States. This data set contains 8 columns that contain the following variables (source:
http://www.boxofficemojo.com):
CI Rank
C2 Movie title
B4 Appendix B Explanation of Data Sets
C3 Name of studio that produced the film
C4 Gross revenue during entire theater release period
C5 Number of theaters that showed the film during release period
C6 Gross revenue during first week of theater release
C7 Number of theaters that showed the film during first week of theater release
C8 Length of release period (in days)
Data Set VII: Standard & Poor's
100 Index Data
This data set contains trading and value information on the 100 stocks in the Standard & Poor's
100 Index as of Friday, August 28, 2009. This data set has 10 columns that contain the fol-
lowing variables (source: http://finance.yahoo.com):
CI Company's stock exchange symbol
C2 Company name
C3 Company's economic sector (e.g., manufacturing)
C4 Stock price at close of business on Thursday, August 27, 2009
C5 Stock price at close of business on Friday, August 28, 2009
C6 Change in stock price from close of business on 8/27/2009 to 8/28/2009
C7 Opening bid for stock price on 8/28/2009
C8 Highest stock price attained on 8/28/2009
C9 Lowest stock price attained on 8/28/2009
C10 Number of shares traded on 8/28/2009
Data Set VIII: McDonald's Data
This data set contains information on the nutritional aspects of McDonald's food. This data set is
reproduced from McDonald's Web site (http//www.mcdonalds.com/usa/eat/nutrition_info.html).
The only alteration involves the approximation of the dietery fiber content of four food items
listed as having less than 1 gram of dietary fiber each, which were all changed to .5 gram. Condi-
ments (ketchup, salad dressing, dipping sauces, and so forth) are not included. This data set has
25 columns that contain the following variables:
CI Menu item
C2 Serving size (in ounces)
C3 Serving size (in grams)
C4 Calories
C5 Calories from fat
C6 Total fat (in grams)
C7 Percent daily value of fat
C8 Saturated fat (in grams)
C9 Percent daily value of saturated fat
C10 Trans fat (in grams)
Cll Cholesterol (in milligrams)
C12 Percent daily value of cholesterol
C13 Sodium (in milligrams)
C14 Percent daily value of sodium
C15 Carbohydrates (in milligrams)
C16 Percent daily value of carbohydrates
C17 Dietary fiber (in grams)
C18 Percent daily value of dietary fiber
Data Set VIII: McDonald's Data B5
C19 Sugars (in grams)
C20 Protein (in grams)
C21 Percent daily value of Vitamin A
C22 Percent daily value of Vitamin C
C23 Percent daily value of Calcium
C24 Percent daily value of Iron
C25 Menu category (e.g., sandwich, non-sandwich chicken, breakfast, and so forth)
Appendix
Statistical Tables
Table I Table of Binomial Probabilities
Table II Values of e" A
Table III Table of Poisson Probabilities
Table IV Standard Normal Distribution Table
Table V The f Distribution Table
Table VI Chi-Square Distribution Table
Table VII The F Distribution Table
Note: The following tables are on the Web site of the text along with Chapters 14 and 15.
Table VIII Critical Values of X for the Sign Test
Table IX Critical Values of T for the Wilcoxon Signed-Rank Test
Table X Critical Values of T for the Wilcoxon Rank Sum Test
Table XI Critical Values for the Spearman Rho Rank Correlation Coefficient Test
Table XII Critical Values for a Two-Tailed Runs Test with a = .05
CI
Appendix C Statistical Tables
Table I Table of Binomial Probabilities
P
11
A"
fl5
>u?
1
,£A3
30
.JU
■DU
70
80
.Oil
QO
.yj
1
.9500
.9000
.8000
.7000
.6000
.5000
.4000
.3000
.2000
.1000
.0500
1
1
osoo
.UJUU
1 noo
. 1UUU
9 OOO
.ZUUU
^OOO
.ouuu
A OOO
.<+UUU
sooo
.J UUU
fC\(\C\
.ouuu
7 OOO
. /UUU
BOOO
.oUUU
QOOO
.VUUU
GSOO
2
.9025
.8100
.6400
.4900
.3600
.2500
.1600
.0900
.0400
.0100
.0025
1
.0950
.1800
.3200
.4200
.4800
.5000
.4800
.4200
.3200
.1800
.0950
2
.0025
.0100
.0400
.0900
.1600
.2500
.3600
.4900
.6400
.8100
.9025
3
.8574
.7290
.5120
.3430
.2160
.1250
.0640
.0270
.0080
.0010
.0001
1
.1354
.2430
.3840
.4410
.4320
.3750
.2880
.1890
.0960
.0270
.0071
2
.0071
.0270
.0960
.1890
.2880
.3750
.4320
.4410
.3840
.2430
.1354
3
.0001
.0010
.0080
.0270
.0640
.1250
.2160
.3430
.5120
.7290
.8574
A
81 AS
f^ft]
.OJOl
AOQA
940 1
1 9QA
069 S
09 S f.
008 1
.UUo 1
001 ft
0001
OOOO
.uuuu
1
.1715
.2916
.4096
.4116
.3456
.2500
.1536
.0756
.0256
.0036
.0005
2
.0135
.0486
.1536
.2646
.3456
.3750
.3456
.2646
.1536
.0486
.0135
3
.0005
.0036
.0256
.0756
.1536
.2500
.3456
.4116
.4096
.2916
.1715
4
.0000
.0001
.0016
.0081
.0256
.0625
.1296
.2401
.4096
.6561
.8145
J
n
V/
. / / JO
.J7UJ
3977
1 681
. 1 uo 1
0778
,u 1 1
03 1 9
.Uj 1 z
01 09
0094.
.UUZH-
0003
0000
.uuuu
0000
-UUUU
1
1
3980
40Q6
.TV7U
3609
.JUUZ
9S09
1 S(V9
. IJUi
0768
-U / uo
0984.
• UZOH-
0064
.UUU4
000s
.UUU J
0000
-UUUU
z
09 1 a
079Q
.U / Zy
9048
34SA
. J4JO
319^
9304
1 393
1 9
.Uj 1Z
008 1
.UUo 1
OOI 1
.UU1 1
3
.0011
.0081
.0512
.1323
.2304
.3125
.3456
.3087
.2048
.0729
.0214
4
.0000
.0004
.0064
.0283
.0768
.1562
.2592
.3601
.4096
.3281
.2036
5
.0000
.0000
.0003
.0024
.0102
.0312
.0778
.1681
.3277
.5905
.7738
o
7^S 1
S^1 A
Ofl 1
1 1 if
.11/0
0AA7
01 S6
-Ul JO
ooai
0007
0001
OOOO
.UUUU
OOOO
.uuuu
i
i
.ZjZ 1
^Q^9
.J7JZ.
^09 S
1 866
.yjy J 1
01 09
• U 1 uz
001 s
-UU 1 J
0001
.UUUl
0000
-UUUU
i.
O^OS
oqra
.WOT 1
9AS8
^9A1
^110
9^AA
1 ^89
OSQS
01 SA
001 9
.UU 1 z
0001
.UUU 1
J
0091
.UUZ 1
01 AA
OS 1 Q
1 SS9
97AS
.Z / Oj
^1 9S
976S
.Z /Dj
1 SS9
08 1 Q
.U0I7
01 Aft
009 1
.uuz 1
,1
H-
0001
OOI 9
.UU1Z
01 SA
.Ul JH-
1 ^89
9^AA
^110
. J 1 1U
^9A1
9AS8
OQ8A
.\Jj\JD
5
.0000
.0001
.0015
.0102
.0369
.0937
.1866
.3025
.3932
.3543
.2321
6
.0000
.0000
.0001
.0007
.0041
.0156
.0467
.1176
.2621
.5314
.7351
1
u
/17G3
90Q7
.ZUV 1
OG9/I
.Uoz4
0980
.UZoU
no70
.UU /o
-UUlo
0009
.UUUZ
OOOO
.uuuu
OOOO
.UUUU
OOOO
.UUUU
1
.2573
.3720
.3670
.2471
.1306
.0547
.0172
.0036
.0004
.0000
.0000
2
.0406
.1240
.2753
.3177
.2613
.1641
.0774
.0250
.0043
.0002
.0000
3
.0036
.0230
.1147
.2269
.2903
.2734
.1935
.0972
.0287
.0026
.0002
4
.0002
.0026
.0287
.0972
.1935
.2734
.2903
.2269
.1147
.0230
.0036
5
.0000
.0002
.0043
.0250
.0774
.1641
.2613
.3177
.2753
.1240
.0406
6
.0000
.0000
.0004
.0036
.0172
.0547
.1306
.2471
.3670
.3720
.2573
7
.0000
.0000
.0000
.0002
.0016
.0078
.0280
.0824
.2097
.4783
.6983
8
.6634
.4305
.1678
.0576
.0168
.0039
.0007
.0001
.0000
.0000
.0000
1
.2793
.3826
.3355
.1977
.0896
.0312
.0079
.0012
.0001
.0000
.0000
X
2
3
4
5
6
7
8
1
2
3
4
5
6
7
8
9
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
6
7
8
9
Table I Table of Binomial Probabilities
of Binomial Probabilities (continued)
P
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
.0515
.1488
.2936
.2965
.2090
.1094
.0413
.0100
.0011
.0000
.0000
.0054
.0331
.1468
.2541
.2787
.2187
.1239
.0467
.0092
.0004
.0000
.0004
.0046
.0459
.1361
.2322
.2734
.2322
.1361
.0459
.0046
.0004
.0000
.0004
.0092
.0467
.1239
.2187
.2787
.2541
.1468
.0331
.0054
.0000
.0000
.0011
.0100
.0413
.1094
.2090
.2965
.2936
.1488
.0515
.0000
.0000
.0001
.0012
.0079
.0312
.0896
.1977
.3355
.3826
.2793
.0000
.0000
.0000
.0001
.0007
.0039
.0168
.0576
.1678
.4305
.6634
.OjUZ
Jo /4
1 LAI
. 1 J4Z
C\A f\A
.U4U4
.U1U1
nnon
.UUZU
.UUUj
.UUUU
r\r\f\r\
.UUUU
f\r\r\r\
.UUUU
r\r\f\r\
.UUUU
.LyoJ
^£7 A
Jo /4
. JUZU
1 ^A
nAn^
.UOUj
ni 7A
.Ul / o
nn^
.UUj J
nnnA
.UUU4
.uuuu
.uuuu
nnnn
.UUUU
.0629
.1722
.3020
.2668
.1612
.0703
.0212
.0039
.0003
.0000
.0000
.0077
.0446
.1762
.2668
.2508
.1641
.0743
.0210
.0028
.0001
.0000
.0006
.0074
.0661
.1715
.2508
.2461
.1672
.0735
.0165
.0008
.0000
.0000
.0008
.0165
.0735
.1672
.2461
.2508
.1715
.0661
.0074
.0006
.0000
.0001
.0028
.0210
.0743
.1641
.2508
.2668
.1762
.0446
.0077
.0000
.0000
.0003
.0039
.0212
.0703
.1612
.2668
.3020
.1722
.0629
.0000
.0000
.0000
.0004
.0035
.0176
.0605
.1556
.3020
.3874
.2985
.0000
.0000
.0000
.0000
.0003
.0020
.0101
.0404
.1342
.3874
.6302
J4o /
. 1U /4
moo
.UZoZ
.UUoU
nm n
.UU1U
nnn i
.UUUl
.uuuu
r\r\r\r\
.UUUU
f\r\r\r\
.UUUU
r\r\f\r\
.UUUU
JO /4
.Z0o4
19 11
. 1Z1 1
nnos
-UUVo
nm A
.UUIO
.UUUl
.uuuu
.uuuu
nnnn
.uuuu
(XI A A
.U /40
1 QT7
JUZU
.ZjJJ
1 ono
m ha
.U1UO
nm a
.UU14
.UUUl
r\r\r\r\
.UUUU
r\r\f\r\
.uuuu
.0105
.0574
.2013
.2668
.2150
.1172
.0425
.0090
.0008
.0000
.0000
.0010
.0112
.0881
.2001
.2508
.2051
.1115
.0368
.0055
.0001
.0000
.0001
.0015
.0264
.1029
.2007
.2461
.2007
.1029
.0264
.0015
.0001
.0000
.0001
.0055
.0368
.1115
.2051
.2508
.2001
.0881
.0112
.0010
.0000
.0000
.0008
.0090
.0425
.1172
.2150
.2668
.2013
.0574
.0105
.0000
.0000
.0001
.0014
.0106
.0439
.1209
.2335
.3020
.1937
.0746
.0000
.0000
.0000
.0001
.0016
.0098
.0403
.1211
.2684
.3874
.3151
.0000
.0000
.0000
.0000
.0001
.0010
.0060
.0282
.1074
.3487
.5987
. jOoo
.JiJO
.\joJy
n i qs
.Ul70
nn^A
nnn^
.UUUD
.uuuu
.uuuu
.uuuu
.uuuu
nnnn
.uuuu
.DLyj
.JOJJ
.ZjOZ
.\Jy jZ
H9AA
.UZDD
.UU04
.uuu /
.uuuu
.uuuu
.uuuu
nnnn
.uuuu
.0867
.2131
.2953
.1998
.0887
.0269
.0052
.0005
.0000
.0000
.0000
.0137
.0710
.2215
.2568
.1774
.0806
.0234
.0037
.0002
.0000
.0000
.0014
.0158
.1107
.2201
.2365
.1611
.0701
.0173
.0017
.0000
.0000
.0001
.0025
.0388
.1321
.2207
.2256
.1471
.0566
.0097
.0003
.0000
.0000
.0003
.0097
.0566
.1471
.2256
.2207
.1321
.0388
.0025
.0001
.0000
.0000
.0017
.0173
.0701
.1611
.2365
.2201
.1107
.0158
.0014
.0000
.0000
.0002
.0037
.0234
.0806
.1774
.2568
.2215
.0710
.0137
.0000
.0000
.0000
.0005
.0052
.0269
.0887
.1998
.2953
.2131
.0867
X
10
11
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
13
1
2
3
4
5
6
7
8
9
10
C Statistical Tables
t>le of Binomial Probabilities
(continued)
P
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
.0000
.0000
.0000
.0000
.0007
.0054
.0266
.0932
.2362
.3835
.3293
.0000
.0000
.0000
.0000
.0000
.0005
.0036
.0198
.0859
.3138
.5688
.j4U4
7Q7/I
.ZoZ4
.UOo /
m
.Ul jo
nn77
.uuzz
.UUUZ
-UUUU
.UUUU
.UUUU
.UUUU
r\r\f\r\
.UUUU
. J4U
.3 / DO
infill
.zuoz
en 1 9
.U / 1Z
m ia
.Ul /4
.uuzy
nnn 3
.UUUJ
.UUUU
.UUUU
.UUUU
.UUUU
flGGQ
.UVoo
7im
7Q^
. ID / o
n£^o
.uo^y
n i a 1
.UlDl
nn7^
.UUZj
.UUUZ
.UUUU
f\r\r\r\
.UUUU
nnnn
.UUUU
f!1 71
nc^7
.UojZ
.ZjOZ
7TQ7
1 /I 1 Q
. i4iy
.Uj j 1
m 7^
.UlZj
.UU 1 j
nnm
.UUUl
f\r\r\r\
.UUUU
nnnn
.UUUU
nn7 1
.UUZ1
.UZ 1 j
. 1 jZy
7^11
.Z.J 1 1
7108
.Z IZo
1 7HS
. IZUo
.U4ZU
.UU / o
.uuuo
.UUUU
nnnn
.UUUU
.0002
.0038
0532
.1585
2270
.1934
.1009
.0291
.0033
.0000
.0000
.0000
.0005
.0155
.0792
.1766
.2256
.1766
.0792
.0155
.0005
.0000
.0000
.0000
.0033
.0291
.1009
.1934
.2270
.1585
.0532
.0038
.0002
.0000
.0000
.0005
.0078
.0420
.1208
.2128
.2311
.1329
.0213
.0021
.0000
.0000
.0001
.0015
.0125
.0537
.1419
.2397
.2362
.0852
.0173
.0000
.0000
.0000
.0002
.0025
.0161
.0639
.1678
.2835
.2301
.0988
.0000
.0000
.0000
.0000
.0003
.0029
.0174
.0712
.2062
.3766
.3413
.0000
.0000
.0000
.0000
.0000
.0002
.0022
.0138
.0687
.2824
.5404
^ 1 ^
7^A7
.ZJ4Z
n^n
HOQ7
.uuv /
nni ^
.UUl J
.UUUl
.UUUU
.UUUU
.UUUU
.UUUU
nnnn
.UUUU
.JjlZ
1 7C7
n^/in
.Uj4U
m i *x
.Ul 1 j
nm fx
.uuio
nnn i
.UUUl
.UUUU
.UUUU
r\r\r\r\
.UUUU
.UUUU
1 1 no
.Z445
7Acn
.ZOoU
1
.1 JOO
C\A ^'X
.U4j 3
nno^
nni 7
.UU1Z
.UUUl
.UUUU
r\r\r\r\
.UUUU
.UUUU
Cil 1 A
.UZ14
Ajyy 1
.Z4J /
.Zlo 1
1 1 H7
. 1 1 U /
n^zio
.Uj4V
nnA^
.UUOj
.uuuo
.UUUU
.UUUU
nnnn
.UUUU
nn7Q
.UUZo
H777
.UZ / /
. 1 J JJ
.Zjj /
. lo4j
.Uo / j
.UZ4J
.UUo4
nnm
.UUUl
.UUUU
.UUUU
nnm
.UUUj
nn^
.UUjj
.UOV1
.IoUj
77 1 A
.ZZ14
1 ^7 1
.13/1
.UojD
m ai
.U14Z
nn i i
.UU1 1
f\r\r\r\
.UUUU
r\r\f\r\
.UUUU
.0000
.0008
.0230
.1030
.1968
.2095
.1312
.0442
.0058
.0001
.0000
.0000
.0001
.0058
.0442
.1312
.2095
.1968
.1030
.0230
.0008
.0000
.0000
.0000
.0011
.0142
.0656
.1571
.2214
.1803
.0691
.0055
.0003
.0000
.0000
.0001
.0034
.0243
.0873
.1845
.2337
.1535
.0277
.0028
.0000
.0000
.0000
.0006
.0065
.0349
.1107
.2181
.2457
.0997
.0214
.0000
.0000
.0000
.0001
.0012
.0095
.0453
.1388
.2680
.2448
.1109
.0000
.0000
.0000
.0000
.0001
.0016
.0113
.0540
.1787
.3672
.3512
.0000
.0000
.0000
.0000
.0000
.0001
.0013
.0097
.0550
.2542
.5133
A Q77
.45 / /
.ZZOo
n/i/in
.U44U
.UUOo
nnns
.UUUo
.UUUl
.UUUU
.UUUU
.UUUU
.UUUU
.UUUU
.JJ7J
.JJJ7
1
. 1 JJ7
.U4U /
nm^
.UU / J
.uuuy
nnn i
.UUUU
.UUUU
.UUUU
.UUUU
1 77Q
. izzy
O^n
.Zj /U
.ZjUl
1 1 'XA
. 1 1 J4
.VjL l
nn^A
nnn^
-UUUj
.UUUU
.UUUU
.UUUU
r\r\r\r\
-UUUU
.0259
.1142
.2501
.1943
.0845
.0222
.0033
.0002
.0000
.0000
.0000
.0037
.0349
.1720
.2290
.1549
.0611
.0136
.0014
.0000
.0000
.0000
.0004
.0078
.0860
.1963
.2066
.1222
.0408
.0066
.0003
.0000
.0000
.0000
.0013
.0322
.1262
.2066
.1833
.0918
.0232
.0020
.0000
.0000
.0000
.0002
.0092
.0618
.1574
.2095
.1574
.0618
.0092
.0002
.0000
.0000
.0000
.0020
.0232
.0918
.1833
.2066
.1262
.0322
.0013
.0000
.0000
.0000
.0003
.0066
.0408
.1222
.2066
.1963
.0860
.0078
.0004
.0000
.0000
.0000
.0014
.0136
.0611
.1549
.2290
.1720
.0349
.0037
X
11
12
13
14
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
2
Table I Table of Binomial Probabilities
of Binomial Probabilities (continued)
P
(Is
1 O
,£A3
50
70
.Oil
on
.yyr
.yj
.0000
.0000
.0000
.0002
.0033
.0222
.0845
.1943
.2501
.1142
.0259
.0000
.0000
.0000
.0000
.0005
.0056
.0317
.1134
.2501
.2570
.1229
.0000
.0000
.0000
.0000
.0001
.0009
.0073
.0407
.1539
.3559
.3593
.0000
.0000
.0000
.0000
.0000
.0001
.0008
.0068
.0440
.2288
.4877
.4633
.2059
.0352
.0047
.0005
.0000
.0000
.0000
.0000
.0000
.0000
.3658
.3432
.1319
.0305
.0047
.0005
.0000
.0000
.0000
.0000
.0000
.1348
.2669
.2309
.0916
.0219
.0032
.0003
.0000
.0000
.0000
.0000
.0307
.1285
.2501
.1700
.0634
.0139
.0016
.0001
.0000
.0000
.0000
.0049
.0428
.1876
.2186
.1268
.0417
.0074
.0006
.0000
.0000
.0000
.0006
.0105
.1032
.2061
.1859
.0916
.0245
.0030
.0001
.0000
.0000
.0000
.0019
.0430
.1472
.2066
.1527
.0612
.0116
.0007
.0000
.0000
.0000
.0003
.0138
.0811
.1771
.1964
.1181
.0348
.0035
.0000
.0000
.0000
.0000
.0035
.0348
.1181
.1964
.1771
.0811
.0138
.0003
.0000
.0000
.0000
.0007
.0116
.0612
.1527
.2066
.1472
.0430
.0019
.0000
.0000
.0000
.0001
.0030
.0245
.0916
.1859
.2061
.1032
.0105
.0006
.0000
.0000
.0000
.0006
.0074
.0417
.1268
.2186
.1876
.0428
.0049
.0000
.0000
.0000
.0001
.0016
.0139
.0634
.1700
.2501
.1285
.0307
.0000
.0000
.0000
.0000
.0003
.0032
.0219
.0916
.2309
.2669
.1348
.0000
.0000
.0000
.0000
.0000
.0005
.0047
.0305
.1319
.3432
.3658
.0000
.0000
.0000
.0000
.0000
.0000
.0005
.0047
.0352
.2059
.4633
.4401
.1853
.0281
.0033
.0003
.0000
.0000
.0000
.0000
.0000
.0000
.3706
.3294
.1126
.0228
.0030
.0002
.0000
.0000
.0000
.0000
.0000
.1463
.2745
.2111
.0732
.0150
.0018
.0001
.0000
.0000
.0000
.0000
.0359
.1423
.2463
.1465
.0468
.0085
.0008
.0000
.0000
.0000
.0000
.0061
.0514
.2001
.2040
.1014
.0278
.0040
.0002
.0000
.0000
.0000
.0008
.0137
.1201
.2099
.1623
.0667
.0142
.0013
.0000
.0000
.0000
.0001
.0028
.0550
.1649
.1983
.1222
.0392
.0056
.0002
.0000
.0000
.0000
.0004
.0197
.1010
.1889
.1746
.0840
.0185
.0012
.0000
.0000
.0000
.0001
.0055
.0487
.1417
.1964
.1417
.0487
.0055
.0001
.0000
.0000
.0000
.0012
.0185
.0840
.1746
.1889
.1010
.0197
.0004
.0000
.0000
.0000
.0002
.0056
.0392
.1222
.1983
.1649
.0550
.0028
.0001
.0000
.0000
.0000
.0013
.0142
.0666
.1623
.2099
.1201
.0137
.0008
.0000
.0000
.0000
.0002
.0040
.0278
.1014
.2040
.2001
.0514
.0061
.0000
.0000
.0000
.0000
.0008
.0085
.0468
.1465
.2463
.1423
.0359
.0000
.0000
.0000
.0000
.0001
.0018
.0150
.0732
.2111
.2745
.1463
.0000
.0000
.0000
.0000
.0000
.0002
.0030
.0228
.1126
.3294
.3706
.0000
.0000
.0000
.0000
.0000
.0000
.0003
.0033
.0281
.1853
.4401
.4181
.1668
.0225
.0023
.0002
.0000
.0000
.0000
.0000
.0000
.0000
.3741
.3150
.0957
.0169
.0019
.0001
.0000
.0000
.0000
.0000
.0000
.1575
.2800
.1914
.0581
.0102
.0010
.0001
.0000
.0000
.0000
.0000
X
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
1
2
3
4
C Statistical Tables
t>le of Binomial Probabilities
(continued)
P
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
.0415
.1556
.2393
.1245
.0341
.0052
.0004
.0000
.0000
.0000
.0000
.0076
.0605
.2093
.1868
.0796
.0182
.0021
.0001
.0000
.0000
.0000
.0010
.0175
.1361
.2081
.1379
.0472
.0081
.0006
.0000
.0000
.0000
.0001
.0039
.0680
.1784
.1839
.0944
.0242
.0026
.0001
.0000
.0000
.0000
.0007
.0267
.1201
.1927
.1484
.0571
.0095
.0004
.0000
.0000
.0000
.0001
.0084
.0644
.1606
.1855
.1070
.0276
.0021
.0000
.0000
.0000
.0000
.0021
.0276
.1070
.1855
.1606
.0644
.0084
.0001
.0000
.0000
.0000
.0004
.0095
.0571
.1484
.1927
.1201
.0267
.0007
.0000
.0000
.0000
.0001
.0026
.0242
.0944
.1839
.1784
.0680
.0039
.0001
.0000
.0000
.0000
.0006
.0081
.0472
.1379
.2081
.1361
.0175
.0010
.0000
.0000
.0000
.0001
.0021
.0182
.0796
.1868
.2093
.0605
.0076
.0000
.0000
.0000
.0000
.0004
.0052
.0341
.1245
.2393
.1556
.0415
.0000
.0000
.0000
.0000
.0001
.0010
.0102
.0581
.1914
.2800
.1575
.0000
.0000
.0000
.0000
.0000
.0001
.0019
.0169
.0957
.3150
.3741
.0000
.0000
.0000
.0000
.0000
.0000
.0002
.0023
.0225
.1668
.4181
.3972
.1501
.0180
.0016
.0001
.0000
.0000
.0000
.0000
.0000
.0000
.3763
.3002
.0811
.0126
.0012
.0001
.0000
.0000
.0000
.0000
.0000
.1683
.2835
.1723
.0458
.0069
.0006
.0000
.0000
.0000
.0000
.0000
.0473
.1680
.2297
.1046
.0246
.0031
.0002
.0000
.0000
.0000
.0000
.0093
.0700
.2153
.1681
.0614
.0117
.0011
.0000
.0000
.0000
.0000
.0014
.0218
.1507
.2017
.1146
.0327
.0045
.0002
.0000
.0000
.0000
.0002
.0052
.0816
.1873
.1655
.0708
.0145
.0012
.0000
.0000
.0000
.0000
.0010
.0350
.1376
.1892
.1214
.0374
.0046
.0001
.0000
.0000
.0000
.0002
.0120
.0811
.1734
.1669
.0771
.0149
.0008
.0000
.0000
.0000
.0000
.0033
.0386
.1284
.1855
.1284
.0386
.0033
.0000
.0000
.0000
.0000
.0008
.0149
.0771
.1669
.1734
.0811
.0120
.0002
.0000
.0000
.0000
.0001
.0046
.0374
.1214
.1892
.1376
.0350
.0010
.0000
.0000
.0000
.0000
.0012
.0145
.0708
.1655
.1873
.0816
.0052
.0002
.0000
.0000
.0000
.0002
.0045
.0327
.1146
.2017
.1507
.0218
.0014
.0000
.0000
.0000
.0000
.0011
.0117
.0614
.1681
.2153
.0700
.0093
.0000
.0000
.0000
.0000
.0002
.0031
.0246
.1046
.2297
.1680
.0473
.0000
.0000
.0000
.0000
.0000
.0006
.0069
.0458
.1723
.2835
.1683
.0000
.0000
.0000
.0000
.0000
.0001
.0012
.0126
.0811
.3002
.3763
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0016
.0180
.1501
.3972
.3774
.1351
.0144
.0011
.0001
.0000
.0000
.0000
.0000
.0000
.0000
.3774
.2852
.0685
.0093
.0008
.0000
.0000
.0000
.0000
.0000
.0000
.1787
.2852
.1540
.0358
.0046
.0003
.0000
.0000
.0000
.0000
.0000
.0533
.1796
.2182
.0869
.0175
.0018
.0001
.0000
.0000
.0000
.0000
.0112
.0798
.2182
.1491
.0467
.0074
.0005
.0000
.0000
.0000
.0000
X
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
1
2
Table I Table of Binomial Probabilities
of Binomial Probabilities (continued)
P
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
nm s
.UUl o
mAA
.UZDO
1 A1A
. lOjO
1 Q 1 A
. ly 10
noil
m77
.UZZZ
nmzi
.UUZ4
nnm
.UUUl
.UUUU
.uuuu
nnnn
.uuuu
nnm
.uuuz
nn aq
no^
1 Q 1 A
.iy lo
1/1^1
. 140 1
n^ 1 s
.Uj lo
nnc^
.UUoJ
nnn^
.UUUj
r\r\f\r\
.UUUU
f\r\r\r\
.UUUU
.uuuu
.uuuu
nn i a
.UUl 4
. 1 jZj
1 707
. 1 / y I
noA 1
.U^Ol
mi7
.UZj /
nm7
.uuzz
.uuuu
.uuuu
nnnn
.uuuu
.uuuu
.uuuz
ni aa
HQS 1
.UVo 1
1 707
. 1 / y I
1 AAO
. 144Z
n^i7
.Uj jz
nn77
.uu / /
nnni
.UUUJ
.uuuu
nnnn
.uuuu
.uuuu
.uuuu
nn^ i
.UUjl
n^ 1 a
.UJ 14
1 A A/1
. 1404
1 7A7
. 1 /OZ
nQ7A
.yjy /0
noon
.uzzu
nn 1 1
.UUl J
.UUUU
.uuuu
.uuuu
.uuuu
nm i
.UUl J
mon
.uzzu
nQ7A
.\Jy /o
1 7A7
. 1 /OZ
1 AfxA
. 1404
n^ 1 a
.UJ 14
nn^ 1
.UUJ 1
.uuuu
nnnn
.uuuu
nnnn
.uuuu
nnnn
.UUUU
nnni
.UUUj
nn77
.UU / /
n^i7
.UjjZ
1/1/17
. 144Z
1 7Q7
A ly 1
noc 1
m aa
.UlOO
nnm
.UUUZ
r\r\f\r\
.uuuu
.0000
.0000
.0000
.0022
.0237
.0961
.1797
.1525
.0443
.0014
.0000
.0000
.0000
.0000
.0005
.0085
.0518
.1451
.1916
.0955
.0069
.0002
.0000
.0000
.0000
.0001
.0024
.0222
.0933
.1916
.1636
.0266
.0018
.0000
.0000
.0000
.0000
.0005
.0074
.0467
.1491
.2182
.0798
.0112
.0000
.0000
.0000
.0000
.0001
.0018
.0175
.0869
.2182
.1796
.0533
.0000
.0000
.0000
.0000
.0000
.0003
.0046
.0358
.1540
.2852
.1787
.0000
.0000
.0000
.0000
.0000
.0000
.0008
.0093
.0685
.2852
.3774
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0011
.0144
.1351
.3774
Jjoj
1 7 1 A
. 1Z10
m 1 ^
.U 1 1 J
nnns
.UUUo
.uuuu
.uuuu
-UUUU
.UUUU
.UUUU
.uuuu
nnnn
.uuuu
777/1
.J 1 /4
.z /uz
n^7A
.Uj /o
nnAG
.UUOo
nnn^
.UUUj
r\f\r\r\
.UUUU
.UUUU
.UUUU
r\r\f\r\
.UUUU
f\r\r\r\
.UUUU
r\r\f\r\
.uuuu
1 QG7
.ZojZ
1 1AQ
n77C
.UZ /o
nm 1
.UUjI
nnno
.UUUZ
.UUUU
.UUUU
r\r\f\r\
.UUUU
.UUUU
r\r\f\r\
.UUUU
n^QA
.UJVO
1 QH1
. iyui
.ZUJ4
H7 1 A
.U / 10
m 7i
.U1ZJ
nm 1
.UUl 1
.uuuu
.uuuu
.uuuu
.uuuu
nnnn
.uuuu
m 11
.U 1 j j
.UoVo
7 1 G7
.Zl oZ
i in/i
. 1 JU4
m^n
.Uj jU
nn/i a
.UU40
nnni
.UUUj
.uuuu
.uuuu
r\r\r\r\
.uuuu
.uuuu
nnoo
.UUZZ
m i o
1 7/1 A
. 1 /40
1 7QO
.1 toy
(Y1A A
.U/40
m a o
.U14o
nm i
.UUl j
.uuuu
.uuuu
r\r\r\r\
.UUUU
.UUUU
.UUUJ
nnso
.UU07
i no i
. 1UV1
1 Q1 A
. 1 y 10
1 1AA
. 1Z44
ni7n
.UJ /u
nn/LQ
.UU4V
nnm
.uuuz
.uuuu
.uuuu
nnnn
.uuuu
.uuuu
.uuzu
.Uj4j
1 A/1 1
.104j
1 A^Q
. IOjV
n7iQ
.U / jy
m a a
.U140
nm n
.UU1U
r\r\f\r\
.UUUU
.UUUU
r\r\f\r\
.uuuu
nnnn
.UUUU
nnn/i
.UUU4
n777
.UZZZ
11/1/1
. 1 144
1 7Q7
. 1 ly 1
1 7m
. 1ZU1
ni^^
.Uj j j
nnio
.uujy
nnm
.UUUl
f\r\r\r\
.UUUU
.UUUU
.uuuu
.UUUl
nn7A
.UU fH
.U0J4
1 ^07
. 1 Jy I
i Am
. 10UZ
m 1 n
.u / 1U
m 7n
.U1ZU
nnn^
.UUUJ
.uuuu
nnnn
.uuuu
.uuuu
.UUUU
nnon
.UUZU
nine
.UjUo
1171
.11/1
1 7A7
. 1 /OZ
1171
.11/1
nine
.UjUo
nnon
.UUZU
r\r\r\r\
.uuuu
r\r\f\r\
.uuuu
noon
.uuuu
.UUUU
nnn^
.UUUJ
n i on
.U1ZU
n7 1 n
.u / 1U
i Am
. 10UZ
1 ^07
. 1 Dy 1
nA^A
.U0J4
nn7d
.UU /4
nnm
.UUUl
nnnn
.uuuu
noon
.uuuu
.UUUU
nnm
.UUUl
nnio
ni^
1 7m
. 1ZU1
1 707
. 1 ly 1
1 1 AA
. 1 144
m77
.UZZZ
nnnA
.UUU4
nnnn
.uuuu
.0000
.0000
.0000
.0010
.0146
.0739
.1659
.1643
.0545
.0020
.0000
.0000
.0000
.0000
.0002
.0049
.0370
.1244
.1916
.1091
.0089
.0003
.0000
.0000
.0000
.0000
.0013
.0148
.0746
.1789
.1746
.0319
.0022
.0000
.0000
.0000
.0000
.0003
.0046
.0350
.1304
.2182
.0898
.0133
.0000
.0000
.0000
.0000
.0000
.0011
.0123
.0716
.2054
.1901
.0596
.0000
.0000
.0000
.0000
.0000
.0002
.0031
.0278
.1369
.2852
.1887
.0000
.0000
.0000
.0000
.0000
.0000
.0005
.0068
.0576
.2702
.3774
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0008
.0115
.1216
.3585
.3406
.1094
.0092
.0006
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.3764
.2553
.0484
.0050
.0003
.0000
.0000
.0000
.0000
.0000
.0000
.1981
.2837
.1211
.0215
.0020
.0001
.0000
.0000
.0000
.0000
.0000
X
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
C Statistical Tables
t>le of Binomial Probabilities
(continued)
P
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
.0660
.1996
.1917
.0585
.0086
.0006
.0000
.0000
.0000
.0000
.0000
.0156
.0998
.2156
.1128
.0259
.0029
.0001
.0000
.0000
.0000
.0000
.0028
.0377
.1833
.1643
.0588
.0097
.0007
.0000
.0000
.0000
.0000
.0004
.0112
.1222
.1878
.1045
.0259
.0027
.0001
.0000
.0000
.0000
.0000
.0027
.0655
.1725
.1493
.0554
.0087
.0005
.0000
.0000
.0000
.0000
.0005
.0286
.1294
.1742
.0970
.0229
.0019
.0000
.0000
.0000
.0000
.0001
.0103
.0801
.1677
.1402
.0497
.0063
.0002
.0000
.0000
.0000
.0000
.0031
.0412
.1342
.1682
.0895
.0176
.0008
.0000
.0000
.0000
.0000
.0008
.0176
.0895
.1682
.1342
.0412
.0031
.0000
.0000
.0000
.0000
.0002
.0063
.0497
.1402
.1677
.0801
.0103
.0001
.0000
.0000
.0000
.0000
.0019
.0229
.0970
.1742
.1294
.0286
.0005
.0000
.0000
.0000
.0000
.0005
.0087
.0554
.1493
.1725
.0655
.0027
.0000
.0000
.0000
.0000
.0001
.0027
.0259
.1045
.1878
.1222
.0112
.0004
.UUUU
r\r\f\r\
.UUUU
f\r\r\r\
.UUUU
r\f\r\r\
.UUUU
.uuu /
nno7
.UUV /
.Ujoo
1 A/1 T
1 CT^
. LOJJ
.\)J 1 1
nmo
.UUZ5
.UUUU
.UUUU
noon
.UUUU
.UUUU
.UUUl
.uuzy
1 1 98
. 1 IZo
.Zl JO
.\}yyo
ni
.Ul JO
.UUUU
f\r\r\r\
.UUUU
r\r\r\r\
.UUUU
.UUUU
r\f\r\f\
.UUUU
nnnA
.uuuo
nne a
.UUoO
.Ujoj
1 Q 1 "7
Ay L /
1 QOA
nAAn
.uoou
.UUUU
.UUUU
r\r\r\r\
.UUUU
.UUUU
.UUUU
nnm
.UUUl
nnon
.UUZU
no 1 ^
1011
. 1Z1 1
.Zoo /
1 QQ 1
.UUUU
.UUUU
.UUUU
.UUUU
.UUUU
noon
.UUUU
nnn^
.UUU J
nn^n
.UUDU
(\ASIA
.U < +o i f
.ZD J J
^7A4
.J /04
.UUUU
r\r\f\r\
.UUUU
.UUUU
r\f\r\r\
.UUUU
.UUUU
nnnn
.UUUU
r\r\r\f\
.UUUU
nnnA
.UUUO
nnoo
.uuvz
i nQ/i
ia nA
. 34U0
.3235
.0985
.0074
.0004
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.3746
.2407
.0406
.0037
.0002
.0000
.0000
.0000
.0000
.0000
.0000
.2070
.2808
.1065
.0166
.0014
.0001
.0000
.0000
.0000
.0000
.0000
.0726
.2080
.1775
.0474
.0060
.0004
.0000
.0000
.0000
.0000
.0000
.0182
.1098
.2108
.0965
.0190
.0017
.0001
.0000
.0000
.0000
.0000
.0034
.0439
.1898
.1489
.0456
.0063
.0004
.0000
.0000
.0000
.0000
.0005
.0138
.1344
.1808
.0862
.0178
.0015
.0000
.0000
.0000
.0000
.0001
.0035
.0768
.1771
.1314
.0407
.0051
.0002
.0000
.0000
.0000
.0000
.0007
.0360
.1423
.1642
.0762
.0144
.0009
.0000
.0000
.0000
.0000
.0001
.0140
.0949
.1703
.1186
.0336
.0032
.0001
.0000
.0000
.0000
.0000
.0046
.0529
.1476
.1542
.0656
.0097
.0003
.0000
.0000
.0000
.0000
.0012
.0247
.1073
.1682
.1073
.0247
.0012
.0000
.0000
.0000
.0000
.0003
.0097
.0656
.1542
.1476
.0529
.0046
.0000
.0000
.0000
.0000
.0001
.0032
.0336
.1186
.1703
.0949
.0140
.0001
.0000
.0000
.0000
.0000
.0009
.0144
.0762
.1642
.1423
.0360
.0007
.0000
.0000
.0000
.0000
.0002
.0051
.0407
.1314
.1771
.0768
.0035
.0001
.0000
.0000
.0000
.0000
.0015
.0178
.0862
.1808
.1344
.0138
.0005
.0000
.0000
.0000
.0000
.0004
.0063
.0456
.1489
.1898
.0439
.0034
.0000
.0000
.0000
.0000
.0001
.0017
.0190
.0965
.2108
.1098
.0182
.0000
.0000
.0000
.0000
.0000
.0004
.0060
.0474
.1775
.2080
.0726
X
20
21
22
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
1
2
3
4
5
6
7
8
9
10
11
Table I Table of Binomial Probabilities
of Binomial Probabilities (continued)
P
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
.0000
.0000
.0000
.0000
.0000
.0001
.0014
.0166
.1065
.2808
.2070
.0000
.0000
.0000
.0000
.0000
.0000
.0002
.0037
.0406
.2407
.3746
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0004
.0074
.0985
.3235
.3074
.0886
.0059
.0003
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.3721
.2265
.0339
.0027
.0001
.0000
.0000
.0000
.0000
.0000
.0000
.2154
.2768
.0933
.0127
.0009
.0000
.0000
.0000
.0000
.0000
.0000
.0794
.2153
.1633
.0382
.0041
.0002
.0000
.0000
.0000
.0000
.0000
.0209
.1196
.2042
.0818
.0138
.0011
.0000
.0000
.0000
.0000
.0000
.0042
.0505
.1940
.1332
.0350
.0040
.0002
.0000
.0000
.0000
.0000
.0007
.0168
.1455
.1712
.0700
.0120
.0008
.0000
.0000
.0000
.0000
.0001
.0045
.0883
.1782
.1133
.0292
.0029
.0001
.0000
.0000
.0000
.0000
.0010
.0442
.1527
.1511
.0584
.0088
.0004
.0000
.0000
.0000
.0000
.0002
.0184
.1091
.1679
.0974
.0221
.0016
.0000
.0000
.0000
.0000
.0000
.0064
.0655
.1567
.1364
.0464
.0052
.0001
.0000
.0000
.0000
.0000
.0019
.0332
.1234
.1612
.0823
.0142
.0005
.0000
.0000
.0000
.0000
.0005
.0142
.0823
.1612
.1234
.0332
.0019
.0000
.0000
.0000
.0000
.0001
.0052
.0464
.1364
.1567
.0655
.0064
.0000
.0000
.0000
.0000
.0000
.0016
.0221
.0974
.1679
.1091
.0184
.0002
.0000
.0000
.0000
.0000
.0004
.0088
.0584
.1511
.1527
.0442
.0010
.0000
.0000
.0000
.0000
.0001
.0029
.0292
.1133
.1782
.0883
.0045
.0001
.0000
.0000
.0000
.0000
.0008
.0120
.0700
.1712
.1455
.0168
.0007
.0000
.0000
.0000
.0000
.0002
.0040
.0350
.1332
.1940
.0505
.0042
.0000
.0000
.0000
.0000
.0000
.0011
.0138
.0818
.2042
.1196
.0209
.0000
.0000
.0000
.0000
.0000
.0002
.0041
.0382
.1633
.2153
.0794
.0000
.0000
.0000
.0000
.0000
.0000
.0009
.0127
.0933
.2768
.2154
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0027
.0339
.2265
.3721
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0003
.0059
.0886
.3074
.2920
.0798
.0047
.0002
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.3688
.2127
.0283
.0020
.0001
.0000
.0000
.0000
.0000
.0000
.0000
.2232
.2718
.0815
.0097
.0006
.0000
.0000
.0000
.0000
.0000
.0000
.0862
.2215
.1493
.0305
.0028
.0001
.0000
.0000
.0000
.0000
.0000
.0238
.1292
.1960
.0687
.0099
.0006
.0000
.0000
.0000
.0000
.0000
.0050
.0574
.1960
.1177
.0265
.0025
.0001
.0000
.0000
.0000
.0000
.0008
.0202
.1552
.1598
.0560
.0080
.0004
.0000
.0000
.0000
.0000
.0001
.0058
.0998
.1761
.0960
.0206
.0017
.0000
.0000
.0000
.0000
.0000
.0014
.0530
.1604
.1360
.0438
.0053
.0002
.0000
.0000
.0000
.0000
.0003
.0236
.1222
.1612
.0779
.0141
.0008
.0000
.0000
.0000
.0000
.0000
.0088
.0785
.1612
.1169
.0318
.0026
.0000
.0000
.0000
.0000
.0000
.0028
.0428
.1367
.1488
.0608
.0079
.0002
.0000
.0000
X
12
13
14
15
16
17
18
19
20
21
22
23
24
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
C Statistical Tables
le of Binomial Probabilities
(continued)
P
.05
.10
.20
.30
.40
.50
.60
.70
.80
.90
.95
.0000
.0000
.0008
.0199
.0988
.1612
.0988
.0199
.0008
.0000
.0000
.0000
.0000
.0002
.0079
.0608
.1488
.1367
.0428
.0028
.0000
.0000
.0000
.0000
.0000
.0026
.0318
.1169
.1612
.0785
.0088
.0000
.0000
.0000
.0000
.0000
.0008
.0141
.0779
.1612
.1222
.0236
.0003
.0000
.0000
.0000
.0000
.0002
.0053
.0438
.1360
.1604
.0530
.0014
.0000
.0000
.0000
.0000
.0000
.0017
.0206
.0960
.1761
.0998
.0058
.0001
.0000
.0000
.0000
.0000
.0004
.0080
.0560
.1598
.1552
.0202
.0008
.0000
.0000
.0000
.0000
.0001
.0025
.0265
.1177
.1960
.0574
.0050
.0000
.0000
.0000
.0000
.0000
.0006
.0099
.0687
.1960
.1292
.0238
.0000
.0000
.0000
.0000
.0000
.0001
.0028
.0305
.1493
.2215
.0862
.0000
.0000
.0000
.0000
.0000
.0000
.0006
.0097
.0815
.2718
.2232
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0020
.0283
.2127
.3688
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0002
.0047
.0798
.2920
.2774
.0718
.0038
.0001
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.3650
.1994
.0236
.0014
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.2305
.2659
.0708
.0074
.0004
.0000
.0000
.0000
.0000
.0000
.0000
.0930
.2265
.1358
.0243
.0019
.0001
.0000
.0000
.0000
.0000
.0000
.0269
.1384
.1867
.0572
.0071
.0004
.0000
.0000
.0000
.0000
.0000
.0060
.0646
.1960
.1030
.0199
.0016
.0000
.0000
.0000
.0000
.0000
.0010
.0239
.1633
.1472
.0442
.0053
.0002
.0000
.0000
.0000
.0000
.0001
.0072
.1108
.1712
.0800
.0143
.0009
.0000
.0000
.0000
.0000
.0000
.0018
.0623
.1651
.1200
.0322
.0031
.0001
.0000
.0000
.0000
.0000
.0004
.0294
.1336
.1511
.0609
.0088
.0004
.0000
.0000
.0000
.0000
.0001
.0118
.0916
.1612
.0974
.0212
.0013
.0000
.0000
.0000
.0000
.0000
.0040
.0536
.1465
.1328
.0434
.0042
.0001
.0000
.0000
.0000
.0000
.0012
.0268
.1140
.1550
.0760
.0115
.0003
.0000
.0000
.0000
.0000
.0003
.0115
.0760
.1550
.1140
.0268
.0012
.0000
.0000
.0000
.0000
.0001
.0042
.0434
.1328
.1465
.0536
.0040
.0000
.0000
.0000
.0000
.0000
.0013
.0212
.0974
.1612
.0916
.0118
.0001
.0000
.0000
.0000
.0000
.0004
.0088
.0609
.1511
.1336
.0294
.0004
.0000
.0000
.0000
.0000
.0001
.0031
.0322
.1200
.1651
.0623
.0018
.0000
.0000
.0000
.0000
.0000
.0009
.0143
.0800
.1712
.1108
.0072
.0001
.0000
.0000
.0000
.0000
.0002
.0053
.0442
.1472
.1633
.0239
.0010
.0000
.0000
.0000
.0000
.0000
.0016
.0199
.1030
.1960
.0646
.0060
.0000
.0000
.0000
.0000
.0000
.0004
.0071
.0572
.1867
.1384
.0269
.0000
.0000
.0000
.0000
.0000
.0001
.0019
.0243
.1358
.2265
.0930
.0000
.0000
.0000
.0000
.0000
.0000
.0004
.0074
.0708
.2659
.2305
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0014
.0236
.1994
.3650
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0038
.0718
.2774
Table II Values of e" A Cl 1
Table II
Values of e A
A
A
U.U
1 oooooooo
1 .UUUUUUUU
j.y
07 07 /i 1 o i
.UzUz4l5*l
o 1
U. 1
.VU4oj /4Z
i o
4.U
.Ulojl J04
O 1
U.Z
.olo 1 jU fJ
l l
4. 1
.UIOj /ZOo
O "2
U.J
. /4UoloZZ
i j
4.Z
.U14VVJJ0
O 1
U.4
.0 /UjzUUj
4. J
.Ul jjOojO
o c
.dUdj jUod
4.4
XjYLL i 1 j4
o a
U.o
.j4ool 104
4. J
.ui i iuvuu
O 7
u. /
.4yojoJjU
1 A
4.0
n 1 nn ^ 1 o a
.UIUUj 1o4
U.o
.44yjzoyO
A 7
4. /
0,0,0.0,0^9°.
.UUVUVjZo
o o
U.V
.4UOJ0V00
4.6
0000707^
.UUozzy / j
1 .U
TA7Q7G/I/1
-JO/o /V44
1
4.y
007/M A^C
.UU/440J5
1 . J
.j jzo / lUo
<* o
J.U
OOA7T7Q^
.uuo / j /y j
1 .2
.jUI 1V4Z1
J. 1
OOAOOA7^
.UUuUVO / j
i 3
1 , j
777^.^ 1 7Q
.Z / Zj J 1 ly
^ J
J.Z
OO 1 ^^ 1 A^A!
.UUjj IOjO
1 .4
7 /i acoaoa
C 1
J.J
.uu4yy i jy
1 .J
.ZZjl jUIO
J. 4
OO/I ^ 1
.UU4j IOjo
1 A
.ZU IoVOjZ
c c
J. J
00/IOCA77
.UU4UoO / /
1 7
.IoZOojjZ
C A
J.O
nno A070A
.UUjOV /oO
1 Q
l.O
.lojzVooy
C 7
J. /
007 1 /i cm
.UUjj4jy /
1 Q
1 /| QC AQA7
-145 , J05UZ
J . o
007077^^
.UUjUz / jj
Z.U
.1 JJJJjZo
j.y
007710/1 A
.UUZ / jV44
2. J
.1ZZ4j04j
A A
O.U
007/1 7Q7C
.UUZ4 / O / J
Z.Z
i 1 oco^i 1 A
. 1 lUoUj 10
A 1
0. 1
0077/17C7
.UUZZ4Z0 /
Z. j
.1UUZjoo4
A J
o.z
007070/1 Q
.UUZUZV43
7 -1
2.4
nom 1 70c
A 1
O.J
001 OTA^IO
.UUIojOjU
7 s
OQ.70Q^nO
.UoZUojUU
A 1
0.4
001 AAJ ^A!
.UU1001 jO
7 A
Z.O
.U /4Z / jjo
A C
O.J
oo 1 cm /i /i
.UU1 jUj44
J "7
Z. /
0A770CC i
.Uo/zUjjI
A A
0.0
001 QAO^I7
.UU1 jOUj /
J Q
Z.o
oaoq i ooa
.UOUolUUO
A 7
0. /
oo 1 7 ion i
.uuizjuyi
z.v
.UjjUZjZZ
A Q
001 1 1 Q7Q
.UU1 1 1 J /o
j.U
o/i 07Q707
.U4V /o /U/
A O
o.y
OO 1 00770
.UU1UU / ly
^ 1
j. 1
O/I ^0/iG70
.U4JU4VZU
7 o
/.U
OOOQ 11CC
.uuuyi loo
J.Z
o/i 07A770
.U4U / OZZU
7 1
/. 1
OOO0 7 C 1 O
.UUUoZj 1U
J.J
nQACS^ 1 7
.UjOooj 1 /
7 7
/ .Z
0007/1 A^Q
.uuu/40jy
3.4
.03337327
7.3
.00067554
3.5
.03019738
7.4
.00061125
3.6
.02732372
7.5
.00055308
3.7
.02472353
7.6
.00050045
3.8
.02237077
7.7
.00045283
Appendix C Statistical Tables
Table II
Values of e A
(continued)
A
A
"7 Q
1,0
.00U4U9 / j
V.J
.UUUU /4cD
/.y
.0003 /O /4
Q A
V.O
.UUUUO 115
5.U
.00033340
(1 1
V. /
.UUUUO IZo
5. 1
.IMJU3UJJ4
G
.UUUUjj4j
5.2
.OOOz /40J
y.y
.UUUUjUI /
O.J
.UUU246JZ
1U.U
.UUUU404U
Q 1
0.4
.000ZZ48 /
11. U
.UUUU 10 /U
O.J
.UUUzUj4/
Iz.U
.UUUUU014
6.0
nnni S/l 1 1
1 j.U
.UUUUUZZO
6. /
.0UU10OJ9
14. U
.UUUUUUoi
6.5
.UUU1 JO/ J
1 0.U
.UUUUUU j 1
5.V
nnni QA^Q
.UUU1 DOjy
10. U
.UUUUUU 1 1
9.0
.00012341
17.0
.00000004
9.1
.00011167
18.0
.000000015
9.2
.00010104
19.0
.000000006
9.3
.00009142
20.0
.000000002
9.4
.00008272
Table III Table of Poisson Probabilities Cl
Table III Table of Poisson Probabilities
X
0.1
0.2
0.3
0.4
A
0.5
0.6
0.7
0.8
0.9
1.0
.9048
.8187
.7408
.6703
.6065
.5488
.4966
.4493
.4066
.3679
1
.0905
.1637
.2222
.2681
.3033
.3293
.3476
.3595
.3659
.3679
2
.0045
.0164
.0333
.0536
.0758
.0988
.1217
.1438
.1647
.1839
3
.0002
.0011
.0033
.0072
.0126
.0198
.0284
.0383
.0494
.0613
4
.0000
.0001
.0003
.0007
.0016
.0030
.0050
.0077
.0111
.0153
5
.0000
.0000
.0000
.0001
.0002
.0004
.0007
.0012
.0020
.0031
6
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0002
.0003
.0005
7
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
X
1.1
1.2
1.3
1.4
A
1.5
1.6
1.7
1.8
1.9
2.0
.3329
.3012
.2725
.2466
.2231
.2019
.1827
.1653
.1496
.1353
1
.3662
.3614
.3543
.3452
.3347
.3230
.3106
.2975
.2842
.2707
2
.2014
.2169
.2303
.2417
.2510
.2584
.2640
.2678
.2700
.2707
3
.0738
.0867
.0998
.1128
.1255
.1378
.1496
.1607
.1710
.1804
4
.0203
.0260
.0324
.0395
.0471
.0551
.0636
.0723
.0812
.0902
5
.0045
.0062
.0084
.0111
.0141
.0176
.0216
.0260
.0309
.0361
6
.0008
.0012
.0018
.0026
.0035
.0047
.0061
.0078
.0098
.0120
7
.0001
.0002
.0003
.0005
.0008
.0011
.0015
.0020
.0027
.0034
8
.0000
.0000
.0001
.0001
.0001
.0002
.0003
.0005
.0006
.0009
9
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0001
.0001
.0002
X
2.1
2.2
2.3
2.4
A
2.5
2.6
2.7
2.8
2.9
3.0
.1225
.1108
.1003
.0907
.0821
.0743
.0672
.0608
.0550
.0498
1
.2572
.2438
.2306
.2177
.2052
.1931
.1815
.1703
.1596
.1494
2
.2700
.2681
.2652
.2613
.2565
.2510
.2450
.2384
.2314
.2240
3
.1890
.1966
.2033
.2090
.2138
.2176
.2205
.2225
.2237
.2240
4
.0992
.1082
.1169
.1254
.1336
.1414
.1488
.1557
.1622
.1680
5
.0417
.0476
.0538
.0602
.0668
.0735
.0804
.0872
.0940
.1008
6
.0146
.0174
.0206
.0241
.0278
.0319
.0362
.0407
.0455
.0504
7
.0044
.0055
.0068
.0083
.0099
.0118
.0139
.0163
.0188
.0216
8
.0011
.0015
.0019
.0025
.0031
.0038
.0047
.0057
.0068
.0081
9
.0003
.0004
.0005
.0007
.0009
.0011
.0014
.0018
.0022
.0027
10
.0001
.0001
.0001
.0002
.0002
.0003
.0004
.0005
.0006
.0008
11
.0000
.0000
.0000
.0000
.0000
.0001
.0001
.0001
.0002
.0002
12
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
Cl 4 Appendix C Statistical Tables
Table III Table of Poisson Probabilities (continued)
X
3.1
3.2
3.3
3.4
A
3.5
3.6
3.7
3.8
3.9
4.0
.0450
.0408
.0369
.0334
.0302
.0273
.0247
.0224
.0202
.0183
1
.1397
.1304
.1217
.1135
.1057
.0984
.0915
.0850
.0789
.0733
2
.2165
.2087
.2008
.1929
.1850
.1771
.1692
.1615
.1539
.1465
3
.2237
.2226
.2209
.2186
.2158
.2125
.2087
.2046
.2001
.1954
4
.1733
.1781
.1823
.1858
.1888
.1912
.1931
.1944
.1951
.1954
5
.1075
.1140
.1203
.1264
.1322
.1377
.1429
.1477
.1522
.1563
6
.0555
.0608
.0662
.0716
.0771
.0826
.0881
.0936
.0989
.1042
7
.0246
.0278
.0312
.0348
.0385
.0425
.0466
.0508
.0551
.0595
8
.0095
.0111
.0129
.0148
.0169
.0191
.0215
.0241
.0269
.0298
9
.0033
.0040
.0047
.0056
.0066
.0076
.0089
.0102
.0116
.0132
10
.0010
.0013
.0016
.0019
.0023
.0028
.0033
.0039
.0045
.0053
11
.0003
.0004
.0005
.0006
.0007
.0009
.0011
.0013
.0016
.0019
12
.0001
.0001
.0001
.0002
.0002
.0003
.0003
.0004
.0005
.0006
13
.0000
.0000
.0000
.0000
.0001
.0001
.0001
.0001
.0002
.0002
14
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
A
X
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
5.0
.0166
.0150
.0136
.0123
.0111
.0101
.0091
.0082
.0074
.0067
1
.0679
.0630
.0583
.0540
.0500
.0462
.0427
.0395
.0365
.0337
2
.1393
.1323
.1254
.1188
.1125
.1063
.1005
.0948
.0894
.0842
3
.1904
.1852
.1798
.1743
.1687
.1631
.1574
.1517
.1460
.1404
4
.1951
.1944
.1933
.1917
.1898
.1875
.1849
.1820
.1789
.1755
5
.1600
.1633
.1662
.1687
.1708
.1725
.1738
.1747
.1753
.1755
6
.1093
.1143
.1191
.1237
.1281
.1323
.1362
.1398
.1432
.1462
7
.0640
.0686
.0732
.0778
.0824
.0869
.0914
.0959
.1002
.1044
8
.0328
.0360
.0393
.0428
.0463
.0500
.0537
.0575
.0614
.0653
9
.0150
.0168
.0188
.0209
.0232
.0255
.0281
.0307
.0334
.0363
10
.0061
.0071
.0081
.0092
.0104
.0118
.0132
.0147
.0164
.0181
11
.0023
.0027
.0032
.0037
.0043
.0049
.0056
.0064
.0073
.0082
12
.0008
.0009
.0011
.0014
.0016
.0019
.0022
.0026
.0030
.0034
13
.0002
.0003
.0004
.0005
.0006
.0007
.0008
.0009
.0011
.0013
14
.0001
.0001
.0001
.0001
.0002
.0002
.0003
.0003
.0004
.0005
15
.0000
.0000
.0000
.0000
.0001
.0001
.0001
.0001
.0001
.0002
X
5.1
5.2
5.3
5.4
A
5.5
5.6
5.7
5.8
5.9
6.0
.0061
.0055
.0050
.0045
.0041
.0037
.0033
.0030
.0027
.0025
1
.0311
.0287
.0265
.0244
.0225
.0207
.0191
.0176
.0162
.0149
Table III Table of Poisson Probabilities Cl
Table III Table of Poisson Probabilities (continued)
.V
5.1
5.2
5.3
5.4
A
5.5
5.6
5.7
5.8
5.9
6.0
7
Z
070^
.U fyj
07 A A
.U /Ul
OASQ
-UDlo
OS so
.UJoU
.UJ44
OSOQ
OA77
.\j i -t 1 1
OAAA
-i
J
.1 J4o
1 7Q^1
. 1ZVJ
1 71Q
AZjy
1 1 BS
. 1 lOJ
. 1 1 J J
1 OB7
. lUoZ
1 OTT
OQBS
OQTS
OBQ7
4
.1719
.1681
.1641
.1600
.1558
.1515
.1472
.1428
.1383
.1339
C
J
1 7s^i
1 7/1 S
. 1 /4o
1 7/1
. 1 /4U
1 77B
. 1 / Zo
171/1
.1/14
1 AQ7
. lOV /
1 A7S
. Id /o
1 asa
1 A^17
1 AOA
<C
i /ion
.14VU
1 S 1 S
1 S17
1 SSS
1 S7 1
.13/1
1 SB/1
1 SQ/1
. 1 jy4
1 Am
. 10U1
1 AOS
.lOUJ
.lOUO
7
1 osa
1 1 7S
. 1 IZj
1 1 A3
. 1 lOJ
1 700
. izuu
1 7^1/1
. 1Z j4
1 7 A7
. 1ZO /
1 70S
1 "37A
. 1 jZO
1 3ST
. 1JJJ
"1 "377
Q
O
OAQ7
07^ 1
077 1
X) 1 1 1
OS 1 O
.Uo 1U
OS4Q
OSS7
OQ7S
OQA7
OQQS
.U770
1 O^
9
.0392
.0423
.0454
.0486
.0519
.0552
.0586
.0620
.0654
.0688
1 u
.UZUU
noon
.UZZU
OO/I 1
.UZ41
.UZoZ
07SS
OIOG
.Vjj4
.Vjjy
mot
.UJOO
.U41 J
1 1
.U1U4
01 1 A
.Ul 10
.uizy
Ol A'X
.U14j
01 S7
.Ul J /
.Ul / J
ni on
.uzu /
.UZZJ
12
.0039
.0045
.0051
.0058
.0065
.0073
.0082
.0092
.0102
.0113
13
.0015
.0018
.0021
.0024
.0028
.0032
.0036
.0041
.0046
.0052
14
.0006
.0007
.0008
.0009
.0011
.0013
.0015
.0017
.0019
.0022
15
.0002
.0002
.0003
.0003
.0004
.0005
.0006
.0007
.0008
.0009
16
.0001
.0001
.0001
.0001
.0001
.0002
.0002
.0002
.0003
.0003
17
.0000
.0000
.0000
.0000
.0000
.0001
.0001
.0001
.0001
.0001
A
X
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.9
7.0
.0022
.0020
.0018
.0017
.0015
.0014
.0012
.0011
.0010
.0009
1
.0137
.0126
.0116
.0106
.0098
.0090
.0082
.0076
.0070
.0064
2
.0417
.0390
.0364
.0340
.0318
.0296
.0276
.0258
.0240
.0223
3
.0848
.0806
.0765
.0726
.0688
.0652
.0617
.0584
.0552
.0521
4
.1294
.1249
.1205
.1162
.1118
.1076
.1034
.0992
.0952
.0912
5
.1579
.1549
.1519
.1487
.1454
.1420
.1385
.1349
.1314
.1277
6
.1605
.1601
.1595
.1586
.1575
.1562
.1546
.1529
.1511
.1490
7
.1399
.1418
.1435
.1450
.1462
.1472
.1480
.1486
.1489
.1490
8
.1066
.1099
.1130
.1160
.1188
.1215
.1240
.1263
.1284
.1304
9
.0723
.0757
.0791
.0825
.0858
.0891
.0923
.0954
.0985
.1014
10
.0441
.0469
.0498
.0528
.0558
.0588
.0618
.0649
.0679
.0710
11
.0244
.0265
.0285
.0307
.0330
.0353
.0377
.0401
.0426
.0452
12
.0124
.0137
.0150
.0164
.0179
.0194
.0210
.0227
.0245
.0263
13
.0058
.0065
.0073
.0081
.0089
.0099
.0108
.0119
.0130
.0142
14
.0025
.0029
.0033
.0037
.0041
.0046
.0052
.0058
.0064
.0071
15
.0010
.0012
.0014
.0016
.0018
.0020
.0023
.0026
.0029
.0033
16
.0004
.0005
.0005
.0006
.0007
.0008
.0010
.0011
.0013
.0014
17
.0001
.0002
.0002
.0002
.0003
.0003
.0004
.0004
.0005
.0006
18
.0000
.0001
.0001
.0001
.0001
.0001
.0001
.0002
.0002
.0002
19
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0001
.0001
.0001
C16 Appendix C Statistical Tables
Table III Table of Poisson Probabilities (continued)
A
X
7.1
7.2
7.3
7.4
7.5
7.6
7.7
7.8
7.9
8.0
.0008
.0007
.0007
.0006
.0006
.0005
.0005
.0004
.0004
.0003
1
.0059
.0054
.0049
.0045
.0041
.0038
.0035
.0032
.0029
.0027
2
.0208
.0194
.0180
.0167
.0156
.0145
.0134
.0125
.0116
.0107
3
.0492
.0464
.0438
.0413
.0389
.0366
.0345
.0324
.0305
.0286
4
.0874
.0836
.0799
.0764
.0729
.0696
.0663
.0632
.0602
.0573
5
.1241
.1204
.1167
.1130
.1094
.1057
.1021
.0986
.0951
.0916
6
.1468
.1445
.1420
.1394
.1367
.1339
.1311
.1282
.1252
.1221
7
.1489
.1486
.1481
.1474
.1465
.1454
.1442
.1428
.1413
.1396
8
.1321
.1337
.1351
.1363
.1373
.1381
.1388
.1392
.1395
.1396
9
.1042
.1070
.1096
.1121
.1144
.1167
.1187
.1207
.1224
.1241
10
.0740
.0770
.0800
.0829
.0858
.0887
.0914
.0941
.0967
.0993
1 1
.0478
.0504
.0531
.0558
.0585
.0613
.0640
.0667
.0695
.0722
12
.0283
.0303
.0323
.0344
.0366
.0388
.0411
.0434
.0457
.0481
13
.0154
.0168
.0181
.0196
.0211
.0227
.0243
.0260
.0278
.0296
14
.0078
.0086
.0095
.0104
.0113
.0123
.0134
.0145
.0157
.0169
15
.0037
.0041
.0046
.0051
.0057
.0062
.0069
.0075
.0083
.0090
16
.0016
.0019
.0021
.0024
.0026
.0030
.0033
.0037
.0041
.0045
17
.0007
.0008
.0009
.0010
.0012
.0013
.0015
.0017
.0019
.0021
18
.0003
.0003
.0004
.0004
.0005
.0006
.0006
.0007
.0008
.0009
19
.0001
.0001
.0001
.0002
.0002
.0002
.0003
.0003
.0003
.0004
20
.0000
.0000
.0001
.0001
.0001
.0001
.0001
.0001
.0001
.0002
21
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0001
X
8.1
8.2
8.3
8.4
A
8.5
8.6
8.7
8.8
8.9
9.0
.0003
.0003
.0002
.0002
.0002
.0002
.0002
.0002
.0001
.0001
1
.0025
.0023
.0021
.0019
.0017
.0016
.0014
.0013
.0012
.0011
2
.0100
.0092
.0086
.0079
.0074
.0068
.0063
.0058
.0054
.0050
3
.0269
.0252
.0237
.0222
.0208
.0195
.0183
.0171
.0160
.0150
4
.0544
.0517
.0491
.0466
.0443
.0420
.0398
.0377
.0357
.0337
5
.0882
.0849
.0816
.0784
.0752
.0722
.0692
.0663
.0635
.0607
6
.1191
.1160
.1128
.1097
.1066
.1034
.1003
.0972
.0941
.0911
7
.1378
.1358
.1338
.1317
.1294
.1271
.1247
.1222
.1197
.1171
8
.1395
.1392
.1388
.1382
.1375
.1366
.1356
.1344
.1332
.1318
9
.1255
.1269
.1280
.1290
.1299
.1306
.1311
.1315
.1317
.1318
10
.1017
.1040
.1063
.1084
.1104
.1123
.1140
.1157
.1172
.1186
11
.0749
.0775
.0802
.0828
.0853
.0878
.0902
.0925
.0948
.0970
12
.0505
.0530
.0555
.0579
.0604
.0629
.0654
.0679
.0703
.0728
Table III Table of Poisson Probabilities C17
Table III Table of Poisson Probabilities (continued)
X
8.1
8.2
8.3
8.4
A
8.5
8.6
8.7
8.8
8.9
9.0
1 1
.UJ / t
.UJ7J
OA 1 A
OzHR
.UHJO
04^0
.UHJ?
.UH-0 1
0^04
14
.0182
.0196
.0210
.0225
.0240
.0256
.0272
.0289
.0306
.0324
1 J
00Q8
.0107
01 76
01 ^ifi
.0147
01 ^8
01 69
.U1U7
.0182
01 94
1 f,
I u
ooso
oo^s
oo^o
.uu / z,
007Q
00R6
oocn
0101
.U1U1
01 OQ
.U 1 U7
17
.0024
.0026
.0029
.0033
.0036
.0040
.0044
.0048
.0053
.0058
18
.0011
.0012
.0014
.0015
.0017
.0019
.0021
.0024
.0026
.0029
19
.0005
.0005
.0006
.0007
.0008
.0009
.0010
.0011
.0012
.0014
20
.0002
.0002
.0002
.0003
.0003
.0004
.0004
.0005
.0005
.0006
21
.0001
.0001
.0001
.0001
.0001
.0002
.0002
.0002
.0002
.0003
22
.0000
.0000
.0000
.0000
.0001
.0001
.0001
.0001
.0001
.0001
A
X
9.1
9.2
9.3
9.4
9.5
9.6
9.7
9.8
9.9
10
.0001
.0001
.0001
.0001
.0001
.0001
.0001
.0001
.0001
.0000
1
.0010
.0009
.0009
.0008
.0007
.0007
.0006
.0005
.0005
.0005
2
.0046
.0043
.0040
.0037
.0034
.0031
.0029
.0027
.0025
.0023
3
.0140
.0131
.0123
.01 15
.0107
.0100
.0093
.0087
.0081
.0076
4
.0319
.0302
.0285
.0269
.0254
.0240
.0226
.0213
.0201
.0189
5
.0581
.0555
.0530
.0506
.0483
.0460
.0439
.0418
.0398
.0378
6
.0881
.0851
.0822
.0793
.0764
.0736
.0709
.0682
.0656
.0631
7
.1145
.1118
.1091
.1064
.1037
.1010
.0982
.0955
.0928
.0901
8
.1302
.1286
.1269
.1251
.1232
.1212
.1191
.1170
.1148
.1126
9
.1317
.1315
.1311
.1306
.1300
.1293
.1284
.1274
.1263
.1251
10
.1198
.1209
.1219
.1228
.1235
.1241
.1245
.1249
.1250
.1251
1 1
.0991
.1012
.1031
.1049
.1067
.1083
.1098
.1112
.1125
.1137
12
.0752
.0776
.0799
.0822
.0844
.0866
.0888
.0908
.0928
.0948
13
.0526
.0549
.0572
.0594
.0617
.0640
.0662
.0685
.0707
.0729
14
.0342
.0361
.0380
.0399
.0419
.0439
.0459
.0479
.0500
.0521
15
.0208
.0221
.0235
.0250
.0265
.0281
.0297
.0313
.0330
.0347
16
.0118
.0127
.0137
.0147
.0157
.0168
.0180
.0192
.0204
.0217
17
.0063
.0069
.0075
.0081
.0088
.0095
.0103
.0111
.0119
.0128
18
.0032
.0035
.0039
.0042
.0046
.0051
.0055
.0060
.0065
.0071
19
.0015
.0017
.0019
.0021
.0023
.0026
.0028
.0031
.0034
.0037
20
.0007
.0008
.0009
.0010
.0011
.0012
.0014
.0015
.0017
.0019
21
.0003
.0003
.0004
.0004
.0005
.0006
.0006
.0007
.0008
.0009
22
.0001
.0001
.0002
.0002
.0002
.0002
.0003
.0003
.0004
.0004
23
.0000
.0001
.0001
.0001
.0001
.0001
.0001
.0001
.0002
.0002
24
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0001
.0001
C18 Appendix C Statistical Tables
Table III Table of Poisson Probabilities (continued)
-V
1 1
1 7
1 _
1 -I
LJ
II
It
A
1 c
id
AO
17
1Q
17
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
1
.0002
.0001
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
2
.0010
.0004
.0002
.0001
.0000
.0000
.0000
.0000
.0000
.0000
5
.0037
.0018
.0008
.0004
.0002
.0001
.0000
.0000
.0000
.0000
4
.0102
.0053
.0027
.0013
.0006
.0003
.0001
.0001
.0000
.0000
5
ATI A
.0224
.0127
.0070
.0037
.0019
.0010
.0005
.0002
.0001
.0001
6
r\A 1 1
.041 1
.0255
.0152
.0087
.0048
.0026
.0014
.0007
.0004
.0002
7
.0646
.0437
.0281
.0174
.0104
.0060
.0034
.0019
.0010
.0005
o
8
.0888
.0655
.0457
.0304
.0194
.0120
.0072
.0042
.0024
.0013
9
.1085
.0874
.0661
.0473
.0324
.0213
.0135
.0083
.0050
.0029
i n
10
1 i a a
.1 194
.1048
.0859
.0663
.0486
.0341
.0230
.0150
.0095
.0058
i i
1 1
1 i n a
.1 194
.1144
.1015
.0844
.0663
.0496
.0355
.0245
.0164
.0106
12
1 aa a
.1094
.1144
.1099
.0984
.0829
.0661
.0504
.0368
.0259
.0176
1 J
.0926
.1056
.1099
.1060
.0956
.0814
.0658
.0509
.0378
.0271
14
.0728
.0905
.1021
.1060
.1024
.0930
.0800
.0655
.0514
.0387
1 c
L5
acq a
.0534
.0724
.0885
.0989
.1024
.0992
.0906
.0786
.0650
.0516
16
.0367
.0543
.0719
.0866
.0960
.0992
.0963
.0884
.0772
.0646
17
.0237
.0383
.0550
.0713
.0847
.0934
.0963
.0936
.0863
.0760
1 o
18
m Ac
.0145
.0255
.0397
.0554
.0706
.0830
.0909
.0936
.091 1
.0844
19
.0084
.0161
.0272
.0409
.0557
.0699
.0814
.0887
.0911
.0888
20
.0046
.0097
.0177
.0286
.0418
.0559
.0692
.0798
.0866
.0888
21
nfvi a
.0024
.0055
.0109
.0191
.0299
.0426
.0560
.0684
.0783
.0846
22
.0012
.0030
.0065
.0121
.0204
.0310
.0433
.0560
.0676
.0769
23
.0006
.0016
.0037
.0074
.0133
.0216
.0320
.0438
.0559
.0669
24
.0003
.0008
.0020
.0043
.0083
.0144
.0226
.0328
.0442
.0557
25
AAA 1
.0001
.0004
.0010
.0024
.0050
.0092
.0154
.0237
.0336
.0446
26
.0000
.0002
.0005
.0013
.0029
.0057
.0101
.0164
.0246
.0343
27
.0000
.0001
.0002
.0007
.0016
.0034
.0063
.0109
.0173
.0254
28
.0000
.0000
.0001
.0003
.0009
.0019
.0038
.0070
.0117
.0181
29
.0000
.0000
.0001
.0002
.0004
.0011
.0023
.0044
.0077
.0125
3 a
JO
aaaa
.0000
aaaa
.0000
AAAA
.0000
AAA 1
.0001
AAAO
.0002
AAA/i
.0006
AA 1 1
.0013
.0026
A A /I A
.0049
AAOQ
.0083
H
aaaa
.0000
aaaa
.0000
AAAA
.0000
AAAA
.0000
AAA 1
.0001
AAAQ
.0003
AAAT
.000/
AA 1 £
.0015
.0030
.0054
J 2
aaaa
.0000
aaaa
.0000
AAAA
.0000
AAAA
.0000
AAA 1
.0001
AAA 1
.0001
AAA A
.0004
AAAA
.0009
AA 1 O
.0018
AAQ A
.0034
3 J
aaaa
.0000
aaaa
.0000
AAAA
.0000
AAAA
.0000
AAAA
.0000
AAA 1
.0001
AAAT
.0002
.0005
AA 1 A
.0010
nm a
.0020
34
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0002
.0006
.0012
35
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0003
.0007
36
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0002
.0004
37
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
.0002
38
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
39
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0000
.0001
Table IV Standard Normal Distribution Table C19
Table IV Standard Normal Distribution Table
The entries in this table give the
cumulative area under the standard
normal curve to the left of z with the
values of z equal to or negative.
z
.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
-3.4
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0002
-3.3
.0005
.0005
.0005
.0004
.0004
.0004
.0004
.0004
.0004
.0003
-3.2
.0007
.0007
.0006
.0006
.0006
.0006
.0006
.0005
.0005
.0005
-3.1
.0010
.0009
.0009
.0009
.0008
.0008
.0008
.0008
.0007
.0007
-3.0
.0013
.0013
.0013
.0012
.0012
.0011
.0011
.0011
.0010
.0010
-2.9
.0019
.0018
.0018
.0017
.0016
.0016
.0015
.0015
.0014
.0014
-2.8
.0026
.0025
.0024
.0023
.0023
.0022
.0021
.0021
.0020
.0019
-2.7
.0035
.0034
.0033
.0032
.0031
.0030
.0029
.0028
.0027
.0026
-2.6
.0047
.0045
.0044
.0043
.0041
.0040
.0039
.0038
.0037
.0036
-2.5
.0062
.0060
.0059
.0057
.0055
.0054
.0052
.0051
.0049
.0048
-2.4
.0082
.0080
.0078
.0075
.0073
.0071
.0069
.0068
.0066
.0064
-2.3
.0107
.0104
.0102
.0099
.0096
.0094
.0091
.0089
.0087
.0084
-2.2
.0139
.0136
.0132
.0129
.0125
.0122
.0119
.0116
.0113
.0110
-2.1
.0179
.0174
.0170
.0166
.0162
.0158
.0154
.0150
.0146
.0143
-2.0
.0228
.0222
.0217
.0212
.0207
.0202
.0197
.0192
.0188
.0183
-1.9
.0287
.0281
.0274
.0268
.0262
.0256
.0250
.0244
.0239
.0233
-1.8
.0359
.0351
.0344
.0336
.0329
.0322
.0314
.0307
.0301
.0294
-1.7
.0446
.0436
.0427
.0418
.0409
.0401
.0392
.0384
.0375
.0367
-1.6
.0548
.0537
.0526
.0516
.0505
.0495
.0485
.0475
.0465
.0455
-1.5
.0668
.0655
.0643
.0630
.0618
.0606
.0594
.0582
.0571
.0559
-1.4
.0808
.0793
.0778
.0764
.0749
.0735
.0721
.0708
.0694
.0681
-1.3
.0968
.0951
.0934
.0918
.0901
.0885
.0869
.0853
.0838
.0823
-1.2
.1151
.1131
.1112
.1093
.1075
.1056
.1038
.1020
.1003
.0985
-1.1
.1357
.1335
.1314
.1292
.1271
.1251
.1230
.1210
.1190
.1170
-1.0
.1587
.1562
.1539
.1515
.1492
.1469
.1446
.1423
.1401
.1379
-0.9
.1841
.1814
.1788
.1762
.1736
.1711
.1685
.1660
.1635
.1611
-0.8
.2119
.2090
.2061
.2033
.2005
.1977
.1949
.1922
.1894
.1867
-0.7
.2420
.2389
.2358
.2327
.2296
.2266
.2236
.2206
.2177
.2148
-0.6
.2743
.2709
.2676
.2643
.2611
.2578
.2546
.2514
.2483
.2451
-0.5
.3085
.3050
.3015
.2981
.2946
.2912
.2877
.2843
.2810
.2776
-0.4
.3446
.3409
.3372
.3336
.3300
.3264
.3228
.3192
.3156
.3121
-0.3
.3821
.3783
.3745
.3707
.3669
.3632
.3594
.3557
.3520
.3483
-0.2
.4207
.4168
.4129
.4090
.4052
.4013
.3974
.3936
.3897
.3859
-0.1
.4602
.4562
.4522
.4483
.4443
.4404
.4364
.4325
.4286
.4247
0.0
.5000
.4960
.4920
.4880
.4840
.4801
.4761
.4721
.4681
.4641
C20 Appendix C Statistical Tables
Table IV Standard Normal Distribution Table
(continued)
The entries in this table give the
cumulative area under the standard
normal curve to the left of z with the
values of z equal to or positive.
z
.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
0.0
.5000
.5040
.5080
.5120
.5160
.5199
.5239
.5279
.5319
.5359
1
5308
. j J70
5438
5478
5517
5557
.JJ7U
5636
5675
.5714
5753
— 1 1 J J
9
5703
5839
5871
.JO / 1
5010
5048
5087
.J70 /
6096
6064
.6103
.6141
3
U.J
.6179
.6217
6955
6703
6331
6368
6406
6443
6480
6517
.UJ 1 1
0.4
.6554
.6591
.6628
.6664
.6700
.6736
.6772
.6808
.6844
.6879
0.5
.6915
.6950
.6985
.7019
.7054
.7088
.7123
.7157
.7190
.7224
n 6
u.u
7957
. / zj /
7901
7394
. / Jit
7357
. / J J /
7380
. / joy
7499
7454
7486
75 1 7
7540
. / Jt7
7
u. /
7580
.761 1
.7642
7673
. / u / J
.7704
.7734
7764
7704
. / / y^t
7873
7859
u.o
7881
. / oo 1
70i n
. / y iu
7030
. / 7J7
7067
. / 7U /
7005
. / 77J
8093
805 1
8078
.ou /
8 1 06
.0 1UU
81 33
0.9
.8159
.8186
.8212
.8238
.8264
.8289
.8315
.8340
.8365
.8389
1.0
.8413
.8438
.8461
.8485
.8508
.8531
.8554
.8577
.8599
.8621
1.1
8643
86 65
8686
8708
.o / uo
8790
.O / i7
8740
.o / ^ry
8770
.0 / / u
8700
.0 / yyj
8810
8830
.OO JU
1.2
8X4.0
88(S0
.OOU7
8888
.oooo
8007
.07U /
8095
8044
8069
8080
8007
.oyy 1
001 5
i . j
0037
.7U
004.0
0066
0089
.7UOi
0000
.7U77
.711J
01 31
01 47
01 69
01 77
1.4
.9192
.9207
.9222
.9236
.9251
.9265
.9279
.9292
.9306
.9319
1.5
.9332
.9345
.9357
.9370
.9382
.9394
.9406
.9418
.9429
.9441
1 .o
.y^-Oj
QA1A
.yH 1 h
.7JUJ
.yj 1 J
.yjZj
.7JJJ
1 7
.yjj^
.7J / J
.7JOZ
.yjy 1
Q^QQ
.7J77
.7OU0
.70jj
1 S
1 .0
.yon- 1
Q6SA
.7UJD
.7004
QA71
.70 / o
.7O0O
.7D7J
.7D77
.7 /UD
1.9
.9713
.9719
.9726
.9732
.9738
.9744
.9750
.9756
.9761
.9767
2.0
.9772
.9778
.9783
.9788
.9793
.9798
.9803
.9808
.9812
.9817
z.l
.yozo
.yoju
.yaio
.yo4z
.yo4o
.yoju
nee,!
.yoj /
L.L
QG6/1
QG6Q
.yo 1 1
.yo 1 j
QC7S
.yo /o
.yso4
.yoo /
QGQfl
.yoyu
2.3
.9893
.9896
.9898
.9901
.9904
.9906
.9909
.991 1
.9913
.9916
2.4
.9918
.9920
.9922
.9925
.9927
.9929
.9931
.9932
.9934
.9936
2.5
.9938
.9940
.9941
.9943
.9945
.9946
.9948
.9949
.9951
.9952
2.6
.9953
.9955
.9956
.9957
.9959
.9960
.9961
.9962
.9963
.9964
2.7
.9965
.9966
.9967
.9968
.9969
.9970
.9971
.9972
.9973
.9974
2.8
.9974
.9975
.9976
.9977
.9977
.9978
.9979
.9979
.9980
.9981
2.9
.9981
.9982
.9982
.9983
.9984
.9984
.9985
.9985
.9986
.9986
3.0
.9987
.9987
.9987
.9988
.9988
.9989
.9989
.9989
.9990
.9990
3.1
.9990
.9991
.9991
.9991
.9992
.9992
.9992
.9992
.9993
.9993
3.2
.9993
.9993
.9994
.9994
.9994
.9994
.9994
.9995
.9995
.9995
3.3
.9995
.9995
.9995
.9996
.9996
.9996
.9996
.9996
.9996
.9997
3.4
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9998
Table V The f Distribution Table
Table V The f Distribution Table
The entries in this table give the critical values
of f for the specified number of degrees
of freedom and areas in the right tail.
t
Area in the Right Tail Under the t Distribution Curve
if
.10
.05
.025
.01
.005
.001
1
3.078
6.314
12.706
31.821
63.657
318 309
2
1.886
2.920
4.303
6.965
9.925
22.327
3
1.638
2.353
3.182
4.541
5.841
10.215
4
1.533
2.132
2.776
3.747
4.604
7.173
5
1.476
2.015
2.571
3.365
4.032
5.893
6
1.440
1.943
2.447
3.143
3.707
5.208
7
1.415
1.895
2.365
2.998
3.499
4.785
8
1.397
1.860
2.306
2.896
3.355
4.501
9
1.383
1.833
2.262
2.821
3.250
4.297
10
1.372
1.812
2.228
2.764
3.169
4.144
11
1.363
1.796
2.201
2.718
3.106
4.025
12
1.356
1.782
2.179
2.681
3.055
3.930
13
1.350
1.771
2.160
2.650
3.012
3.852
14
1.345
1.761
2.145
2.624
2.977
3.787
15
1.341
1.753
2.131
2.602
2.947
3.733
16
1.337
1.746
2.120
2.583
2.921
3.686
17
1.333
1.740
2.110
2.567
2.898
3.646
18
1.330
1.734
2.101
2.552
2.878
3.610
19
1.328
1.729
2.093
2.539
2.861
3.579
20
1.325
1.725
2.086
2.528
2.845
3.552
21
1.323
1.721
2.080
2.518
2.831
3.527
22
1.321
1.717
2.074
2.508
2.819
3.505
Zj
1 T 1 Q
1 "71/1
z.OUU
Z.oU /
24
1.318
1.711
2.064
2.492
2.797
3.467
25
1.316
1.708
2.060
2.485
2.787
3.450
26
1.315
1.706
2.056
2.479
2.779
3.435
27
1.314
1.703
2.052
2.473
2.771
3.421
28
1.313
1.701
2.048
2.467
2.763
3.408
29
1.311
1.699
2.045
2.462
2.756
3.396
30
1.310
1.697
2.042
2.457
2.750
3.385
31
1.309
1.696
2.040
2.453
2.744
3.375
32
1.309
1.694
2.037
2.449
2.738
3.365
33
1.308
1.692
2.035
2.445
2.733
3.356
34
1.307
1.691
2.032
2.441
2.728
3.348
35
1.306
1.690
2.030
2.438
2.724
3.340
c
Ta
df
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
00
il Tables
The t Distribution Table (continued)
Area in the Right Tail Under the t Distribution Curve
.10
.05
.025
.01
.005
.001
.306
1.688
2.028
2.434
2.719
3.333
.305
1.687
2.026
2.431
2.715
3.326
.304
1.686
2.024
2.429
2.712
3.319
mA
.304
1.085
i nil
2.023
1 A 1£.
2.420
i mo
2. /OS
1 1 1 Q
3.313
.303
1.084
i m 1
2.021
1 All
LAZi
1 Hf\A
2. /04
3.30/
.303
1.683
2.020
2.421
2.701
3.301
.302
1.682
2.018
2.418
2.698
3.296
.302
1.681
2.017
2.416
2.695
3.291
.301
1 /ion
1.O80
i n 1 c
2.U15
1 /I 1 A
2.414
i £ni
2.o92
1 1 OA
3.250
.301
1.0/9
i m a
2.014
2.412
i tnn
2.09U
3.251
.300
1.679
2.013
2.410
2.687
3.277
.300
1.678
2.012
2.408
2.685
3.273
.299
1.677
2.011
2.407
2.682
3.269
inn
1.0/ /
i n 1 n
2.U1U
i a nc
2.405
i Aon
2.05U
3.2o5
inn
l.O/O
i nnn
2.009
o /) nQ
2.4U3
2.0/5
3.2ol
.298
1.675
2.008
2.402
2.676
3.258
.298
1.675
2.007
2.400
2.674
3.255
.298
1.674
2.006
2.399
2.672
3.251
.297
1.674
2.005
2.397
2.670
3.248
.29 /
1.0/3
i nn/i
2.004
1 int
2.390
1 AAO
2.000
3.245
.297
1.673
2.003
2.395
2.667
3.242
.297
1.672
2.002
2.394
2.665
3.239
.296
1.672
2.002
2.392
2.663
3.237
int
1.0/1
i nn 1
2.U01
1 ini
2.391
2.002
1 11 A
3.234
on/;
1.0/ 1
i nnn
2.U0U
1 Qnn
2.390
i AAn
2.000
5. HI
.296
1.670
2.000
2.389
2.659
3.229
.295
1.670
1.999
2.388
2.657
3.227
.295
1.669
1.998
2.387
2.656
3.225
in?
Low
i nno
1.998
1 T OiC
2.300
2. 055
3.223
.295
1.669
1.997
2.385
2.654
3.220
.295
1.668
1.997
2.384
2.652
3.218
.294
1.668
1.996
2.383
2.651
3.216
.294
1.668
1.995
2.382
2.650
3.214
9Q4
1 .uu /
1 .77J
9 ^89
Z.U47
.294
1.667
1.994
2.381
2.648
3.211
.294
1.667
1.994
2.380
2.647
3.209
.293
1.666
1.993
2.379
2.646
3.207
.293
1.666
1.993
2.379
2.645
3.206
.293
1.666
1.993
2.378
2.644
3.204
.293
1.665
1.992
2.377
2.643
3.202
.282
1.645
1.960
2.326
2.576
3.090
Table VI Chi-Square Distribution Table C23
Table VI Chi-Square Distribution Table
The entries in this table give the critical values of x 2 for the specified
number of degrees of freedom and areas in the right tail.
X 2
Area in the Right Tail Under the Chi-square Distribution Curve
df
.995
.990
.975
.950
.900
.100
.050
.025
.010
.005
1
0.000
0.000
0.001
0.004
0.016
2.706
3.841
5.024
6.635
7.879
2
0.010
0.020
0.051
0.103
0.211
4.605
5.991
7.378
9.210
10.597
3
0.072
0.115
0.216
0.352
0.584
6.251
7.815
9.348
11.345
12.838
A
4
O 907
U.ZU /
O 9Q7
ASA
n 7 1 1
U. / 1 1
1 oaa
1 .UD4
7 77Q
/ . / ly
Q ASS
11 1 A%
1 1 . 14 J
1 1 977
1 J.Z / /
1 A SAO
J
n a 1 9
u.4iz
n ^a
U. JJ4
U.o J 1
1 1 A^
I.OIU
Q 9^A
1 1 H7H
1Z.OJJ
1 ^ nsA
1 J.UoO
1 a 7^n
ID. /JU
6
0.676
0.872
1.237
1.635
2.204
10.645
12.592
14.449
16.812
18.548
7
0.989
1.239
1.690
2.167
2.833
12.017
14.067
16.013
18.475
20.278
8
1.344
1.646
2.180
2.733
3.490
13.362
15.507
17.535
20.090
21.955
Q
y
I . / Jj
i nss
Z.Uoo
i 7on
z. /uu
J. jZj
zL 1 AS
4. lOo
1 ZL ASA
14.004
1 A Q1 Q
1 Q 07^
71 AAA
Zl .OOO
73 SSQ
Zj.jo7
1 n
1U
9 1 SA
Z. 1 JO
Z.jjo
7zL7
J.Z4 1
J.V4U
A SA^
1 ^ QS7
1 J.yo 1
1 S "307
lo. JV 1
7H A'A'X
ZU.45J
ZJ.ZUV
7^ 1 SS
Zj. loo
11
2.603
3.053
3.816
4.575
5.578
11.215
19.675
21.920
24.725
26.757
12
3.074
3.571
4.404
5.226
6.304
18.549
21.026
23.337
26.217
28.300
13
3.565
4.107
5.009
5.892
7.042
19.812
22.362
24.736
27.688
29.819
1 ZL
14
zi_ ens.
4.U / J
zL AAO.
4.00U
S, A7Q
J.OZV
A ^7 1
0. J / 1
7 7QO
7 1 OAzl
Zl .U04
OQ AS">
ZJ.OoJ
7A 1 1 Q
ZO. 1 Ly
70 1 A 1
Ly. 141
3. 1 3 1 Q
j 1 . j 1 y
1 c
1 J
zL AO.1
4.0U1
^ 77Q
J.ZZV
A 7A7
D.ZDZ
7 7A1
/ .ZD1
S ">zl_7
77 3.H.7
OA QQA
Z4.WO
77 ZlSS
Z / .400
an i \7S
JU. J /
^.7 sm
OZ.oUl
16
5.142
5.812
6.908
7.962
9.312
23.542
26.296
28.845
32.000
34.267
17
5.697
6.408
7.564
8.672
10.085
24.769
27.587
30.191
33.409
35.718
18
6.265
7.015
8.231
9.390
10.865
25.989
28.869
31.526
34.805
37.156
1 Q
A SzLzl
0.044
S QO.7
1U. 1 1 /
I 1 A^ 1
I I .OJ 1
77 7HA
Z / .ZU4
an 1 zlzl
JU. 144
8^7
^A 1 Q 1
jO. iy 1
ao coo
jo.joZ
7(1
zu
/ ,4J4
£ 7AO.
o.ZOU
y. jy i
lU.o J 1
1 7 zlzL^
1Z.44J
7S A 1 7
Zo.41Z
31 zL 1 n
J 1 ."4-10
a/1 1 7n
j4. 1 /U
"37 ">AA
j / . DOO
T.Q QQ7
jy.yy 1
21
8.034
8.897
10.283
11.591
13.240
29.615
32.671
35.479
38.932
41.401
22
8.643
9.542
10.982
12.338
14.041
30.813
33.924
36.781
40.289
42.796
23
9.260
10.196
11.689
13.091
14.848
32.007
35.172
38.076
41.638
44.181
9A
Z4
Q SSA
in
1U.O JO
1 9 An i
1Z.4U1
1 ^ CAS
1 J.o4o
1 J.OJV
JJ.iyO
JO. 41 J
QAA
A9 QSO
A^ ^^Q
4J.JJV
25
10.520
11.524
13.120
14.611
16.473
34.382
37.652
40.646
44.314
46.928
26
11.160
12.198
13.844
15.379
17.292
35.563
38.885
41.923
45.642
48.290
27
11.808
12.879
14.573
16.151
18.114
36.741
40.113
43.195
46.963
49.645
28
12.461
13.565
15.308
16.928
18.939
37.916
41.337
44.461
48.278
50.993
29
13.121
14.256
16.047
17.708
19.768
39.087
42.557
45.722
49.588
52.336
30
13.787
14.953
16.791
18.493
20.599
40.256
43.773
46.979
50.892
53.672
40
20.707
22.164
24.433
26.509
29.051
51.805
55.758
59.342
63.691
66.766
50
27.991
29.707
32.357
34.764
37.689
63.167
67.505
71.420
76.154
79.490
60
35.534
37.485
40.482
43.188
46.459
74.397
79.082
83.298
88.379
91.952
70
43.275
45.442
48.758
51.739
55.329
85.527
90.531
95.023
100.425
104.215
80
51.172
53.540
57.153
60.391
64.278
96.578
101.879
106.629
112.329
116.321
90
59.196
61.754
65.647
69.126
73.291
107.565
113.145
118.136
124.116
128.299
100
67.328
70.065
74.222
77.929
82.358
118.498
124.342
129.561
135.807
140.169
.2 "S
2 - o
** « in
.E bo 3
« ~ Sb
m m a)
*n -C "c
S .£
|2
3
m
5
>
.a
|2
as oo co as in so >— < >— <
cn in i-H o; h o; q
o\ *o n so in ^t- ^t-
oo in as Tt
co CN
cri \d n o~i S iri "n ^ ^
h - ^ a\
^ (N
os so co os
in oo co
SO CO OS
*o oo i— i in
in as -st
r- in in rj- -sf
h >n i^i 4
H O O rH OO
ci — ; O)
co co m co ci
co co co co
co m co cn co
°4 as so co as h in \f
^ as CN i-H
m co co
in os cn in
o> o in
_o as in so so
^ cn cn sq
C ai ^' ©
SO Os CN ,—| " —|
so r- oo \o
^ cn sq Os ^f;
Os Os l> -^t O
in as cn r— i i— i
cn cn in os
.„ in >— i oo as
^} N h o\ n
Os od in i— h
in Os cn ' — 1 < — 1
O CN O O
g O 00 O CN
O as o od cn
in on cn "—i "—i
m t f co co co co
^ m rf t co co co
CN t t " —| " —|
r-; *t sq
r-^ ^o in in
r- so in in
cn >— < so in
OO SO OO CN OO
f>- so in in *sf
oo (N h m Tf
o\ i — as co Os
t>- so in in *sf
oo >o so in in
^f" xt" CO CO CO
NO CN CN SO CO
■<t cn o oo
^ -t ri rn
Tj" "^J" Tl" CO CO
^ ^ -!t cn
t t
oo so so m, m, ^ -t ^ t
xj- i— i co oo CO
oo H so in in
CN CN CN CN CN
oo cn t co as
Tf" tJ" CO CO CN
CN CN CN CN CN
CNCNCNCNCN CN CN CN CN CN
M CN h h
COCNCNCNCN CNCNCNCNCN CI CN CI >— <
O O CN T}" OO
i-h p as oo
CO CO CN CN CN
so r-~ oo t— i
o as as oo
CO CO CN CN CN
CO CO CO CO CN
CO CO CO CO CO
CO co co cn co
CO CO CO CO CO
CO CO CO CO CO
CO CO CO CO CO
CO CO CO CO CO
CO CO CO CO
CN CN CN CN CN
CN CN CN CN CN
OO CO oo -3- O
oo oo r-~ t~- r-
ci ci ci cn ci
o\ © o as
CO CN i — 1 OO
ci ci ci -h
in t> r-~
(S H 0\
ci ci ci -h
in t> o
in co cn p
ci ci ci cn
COCNCNCNCN CI CI CN CN
COCOCOCOCN CNCNCNCN
CO CO CO CO CO
CO CO CO CO CO
CO CO CO CO CO
cO cO cO cO CO
as -sj- o so
sq in in in
CO CO CO CO CO
CO CO CO
as t> sq ^
ci ci CN ci
00 O O O
as oo r-- in
ci cn ci cn
r- as oo as
O oo o in
co ci ci cn
CO CN CN CN
cn co co CN
in -rf ^t- ^f- -sf
Os oo
co co
COCOCOCOCO COCOCOCN
oo so so in inin^t^t^t ^ ^ co co co co
<n m - — i cN Os
oo p as
as i> t>- sd in
m in m in rf
^ rl-
CO CO CO CO
CO CO CO
as oo r- so so so in in in in
in in in in ^- rf ^j- rt - ^--^t^-co
© as oo oo r- r^sososooi Oj Oi O; "n »n m, m m m m ininin^-
co CN * o
3
d
as as as oo oo
oo oo oo oo oo
oor-r-r-r- f- l> r- so
r- n m \f in sor-ooasO ^HCNco^t-in
oo as ©
Degrees of Freedom for the Denominator
C24
Si °
id £ 13
in jj =
PJ a. E
- "O =
o c «
a -a
c n
in u i-
01 o
s 3 1
■=■*■<»
s r
£ * o
l> C V
3 = «;
(0 — x-
in *- o
Z « in
Z f 8
in at aj
ai
■a
3
O
u
C
o
5
>
o n oo
CO OO rn oo Tf
O "^t p rn
o ai t oo o
' CO 1
^o r- ^: — 1 00
O ©
_^ ^O co no cn
© Tt o ^ rt
o ai i-' oo >o
cn i— i
no on o r-
rH Tt *-h IT) CS
00 OS if CO \o
on cn *
on
en in so co
On "t (S \£)
oo cn >— i
h ^ in n
h 1; rn IT;
*o o\ i - ' oo \d
r-~ cn "— '
Os
t — "d* oo SO
t> cn in on r~-
NO ON *sf" OO NO
IT) CO H
ON
cn cn © oo
rH en CN On
l> ON *d/ ON NO
co co i— i
o rt in m
in © © no *d;
On On nO* O OO
On CO i— i ■— i
t CO t rt O
CO — ■— I T— I
NO
M r- ^ o in
o> tN] h Tf H
■^j- -^j- co co cn
oo oo i— i r~- rt
ON rt 00 (N
■^j- -^j- co CO cn
in rf co co co
in rf co co co
to ^ CO CO CO
•n rf co co
n ^ m m
in ^ ^ co co
•n ri- ri- co co
•n rt - co co
in r|- -^t co
o o co © in
no on r— i oo
in ^- ^- co
in ^j- ^j- co
in in
in in ^
no © t no fr~~
On oo no "n
oi oi oi oi oi
co rt rt rt rt
CO rt rt rt rt
CM no ^ co
i OS OO t~- \o
pi d H H H
no -h OO OO On
rH q 00 h o
co co" rt rt rt'
rt rt rt
cn cn cn rt rt
co oo in in no
*fr rt — j © on
cn cn cn cn rt
cO cO cO
CO CO cO
cn co cn
cn cn co co cn
no oo oo On
i> no co rt
cn cn co co cn
cn cn co co cn
co co co cn
no >n >n ^- ^ t|- m m
© on rt oo cn
no oo © oo
no in in in'^'
m h n in
no 'nI - co rt < — 1
Tl-' Tt
r-NONOinin in in ^ ^
© cn r- rt o
cn cn rt -h
rt rt rt rt rt
-h in © in
Tt co co rt
rt rt rt rt rt
rt rt rt rt rt
t © on in
in in co co
cn rt rt rt rt
rt rt rt rt rt
rt rt rt rt rt
rt rt rt rt rt
rt rt rt rt rt
co i> >— i no rt
on oo oo t~- r-~
rt rt rt rt rt
On rt i> rt r-
On On oo oo r-~
rt rt rt rt rt
rt rt rt rt
m cn cn rt rt
m cn cn cn cn
cn cn cn cn cn
rtrtrtrtrt
rtrtrtrtrt
■ — ' c~~ r < — 1 xj"
go © in rt on
CO 0O h h *C
rt in — i © ©
in ^t- co i rt
NO NO NO NO NO
co co cn cn cn
cn cn cn
on rt NO I NO
no no in in
^ -t
no no in in in
rtrtrtrtrt
—i t~~ -h oo
cn rt rt rt --h
rtrtrtrtrt
so rt On no cn
cn cn rt rt rt
rtrtrtrtrt
rtrtrtrtrt
co © r- ^f- i— i
•n in t}-
rtrtrtrtrt
rtrtrtrtrt
oo in rt On no
\o so so >n >n
rtrtrtrtrt
rtrtrtrtrt
rtrtrtrtrt
rtrtrtrtrt
rtrtrtrtrt
on in rt on c*-
© © © On On
co co* co* rt" rt'
co co cn cn cn
^; cn
co co cn co
cn cn cn cn cn
^ ^j-
in in in >n >n
~ 00 © -d-
© 00 OO NO
rt on cn
' — J On On r~-^
rt rt " —|
rt rt rt >— >
' — 1 On rt 00
rt rt ©
rt rt rt' rt
no cn no rt
cn rt
rt rt rt" rt
— on rt oo
in cn co "— <
rt rt rt" rt
rt rt rt rt
in cn no rt
no in ^ cn
rt' rt' rt rt
in rt in rt
no in
rt' rt' rt" rt
rt rt rt rt
cn © cn ©
© on oo
co" rt* rt rt"
cn cn cn rt
co co cn cn
■^f co cn
in in in in
"— i cn cn -dr in no r— oo on © -h cn cn in noc~^ooon©
Degrees of Freedom for the Denominator
<u a, o
m 5 •=
9 a. E
^ « o
o =
U. (0 "O
> c
w § «
01 u >-
= _ o
_ -ts «
3 I |
S t =
k. LA
** 01 o
S£t
"So
oi -a -a
3 c oi
** — *
» « *
.Sf 3
5
g £
01
T3
111
3
C
O
_0j
&
3
X)
m
5
01
>
XI
|2
=2
00 00 O
oo tj- in r-
i— 5 on od in
in -h
no r-1 in
■«t no r-;
on od in
r-i Tj- r-i 00
t O)
ci 0\ oo iri \f
3 "
O O t O
© ON NO <d-
ON T]- O ON O
OO ©
in m oo o
© on od no
3
o> en
od on od NO
no on oo no rj-
© -h no in
CNJ cn p CN p
© on on \o in
cn i— (
in cn on on
NO CN rH cn -h
on on no in
cN — i
ri
NO 00 ON -H
^ "1 ^ "1 t
in on on no in
o in on
■n o in on r-^
On On On NO "n
ON
^ in h h \q
> — 3 od c5 t ■ no"
^O — I -H
— t> t> NO ON
fr~~ cn | on r-; in
rn rn ri ri (N
CN CN O -d-
cn O oo no
cn cn CN CN
m m m cn CN
m m m cn CN
cn cn cn cn CN
oo -d;
cn cn cn
cn cn cn cn cn
O t> 00 o >— '
p in cn p ON
tJ- cn cn cn ri
cn cn cn cn
no in -d" oo
o cn on
cn cn cn cn
cn cn cn cn
h ^ oj O
cn" cn cn cn
cn cn cn cn
cn oo in cn cn
^ cn cn cn cn
r*~i r*~i r*~i r*~i
•*t cn cn cn
NO in t— NO rH
o
^ ^ ^
Tt Tt NO NO O
H t-; CN i-H
in Tf Tf
in <n in in Tt
no in no on cn
t ^ ^ H . "1
CN CN CN CN CN
CN CN CN CN ri
cn cn [~~~ ©
in cn cN CN
ri ri ri ri ri
r> r> oo •— i in
in ^ m m cn
ri ri ri ri ri
O O OO
no in cn cn
ri ri ri ri ri
CN CN cn cn n
n cn cn cn n
CN CN CN CN CN
rj rj
OO l>
ri ri cn cn cn
in in f- o
oo no no in
ri ri ri ri ri
o o in on
on oo r — no in
ri ri ri ri ri
in in r- o
ON OO [~~~ [~~~
ri ri cn cn
m CN CN CN CN
m m cN CN CN
© -h cn no o
CN — h p On On
cn cn cn ri ri
cn cn cn cn cn
m m cn cn cn
m m cn cn cn
^
h M oo -h ooincNOoo ooncnon
ooononon oo oo oo oo r- o in in m
ri ri rA rA rA rA rA rA i-J i-J *-5 *-5 — 3 — 3
rl CO o t — ^t — OOnO^ nonoooo
© © © ON ONONOOOOOO l>- NO NO '^f
ri ri ri ri -h r-5 r-3 tA tA tA ~A ~A ~A ~A
inONOcnON no ^ o\ h> on on m cn
"-H"-h©oon onononoooo o- no no »n
ri ri ri ri -h rA rA rA rA rA ~A ~A ~A ~A
o\ m h b t >— r-NO-d/cN t t o\ f*
^h^h^hoo oonononon oor^NOin
ririririri csj^_:^-:^-:^-: ^^^^
m oo t - 1 h in (N O t — no oooocnri
cn p pppONON oo r~~- r~~- no
ririririri ri ri ri h h
oo cn on no cn
CN CN — ; — ; — ;
ri ri ri ri ri
in r- cn
cn cn cN cN
ri ri ri ri
CN OO rH OO
tJ- cn cn cn cN
ri ri ri ri ri
CN CN CN CN CN
on in -h oo in
cn cn
ri ri ri ri ri
CN CN CN CN CN
CN CN CN CN CN
CN CN CN CN CN
CN CN CN CN CN
in h h ^ -
oo oo t — t — t —
cn cn ci ri ri
— ^ no cn o r —
o on on on oo
cn ri ri ri ri
m m cn cn cn
m m cn cn cn
on in -h oo in
Tj" Tf rj- rn m
Tj-" Tf
o r~- in cn <— »
— ; p p p p
CN CN CN CN CN
CN CN CN CN CN
CN CN CN
ri ri ri
oo NO ri o
CN CN CN CN CN
ri ri ri ri ri
cn cn cn CN
ri ri ri ri
CN CN CN CN CN
CN CN CN CN CN
ON NO ^ CN O
Tt
ri ri ri ri ri
rj CN CN CN CN
OO NO CN ©
^O ^O nO nO nO
ri ri ri ri ri
■d" CN © OO nO
OO OO 0C h h
ri ri ri ri ri
m cn cn cn cn
t — "d" CN ©
cn cn cn cn
CN © OO NO "d"
cn cn cn CN CN
Tt Tt Tt
CN CN -h -h
2 3
CN CN -h -h
no oo cn cn
i — J © © ON
ri ri ri ~A
ri r- r-
CN < — ^ O ON
ri ri ri ~A
o oo cn cn
CN ©
ri ri ri ri
CN CN CN CN
CN CN CN CN
cn in © i— i
>n ^ cn
ri ri ri ri
ON -H NO NO
no no in ^
ri ri ri ri
<N Tt ON ©
ON OO [~~~ [~~~
ri ri ri ri
cn cn cn cn
—i © ©
Tt" Tt"
h n n 't in Nor~-oooN© >— i cS cn m
00 ON ©
© © © ©
cn in ©
Degrees of Freedom for the Denominator
C26
nj -a *i
<u <ii o
o 5 .s
T Q- E
81 o
o -o =
m- £ gj
U. ro "O
O > c
ui § "0
cu g b
3 _ o
lis
— t; *
» 3 E
01
=5 I
01 o
60
. E
01 o
0) -o -o
3 c oi
5 = 2
J a ■s
.E M qJ
"> *™ bo
0) V (Si
*n j: -a
s =
01
T3
01
o
u
C
O
a
01
>
ai
i— < oo oo m
O i-h f; H
co in m m
t ^ M h
ri iri rn rn
cn r- so o so
in ^ x h
cN os in rn en
so so r-~ cn i>
tSj H oo H
cN as in rn rn
t t co H
t-^ -sf --; i 0O CN
-h as in en cn
0O CN
rn rn
CN cn o
CN -sf CN
r-5 as in
1 ' CN
h (S
o as in
o cN - — ' oo
■^f CN OS CN
as in rn rn
as in m m
as as in m m
as as in m m
— h in t co t~~-
as m cN as m
oo as in rn m"
O m oo
CN rn cN
oo as in
c — as <n m
oo cN rn ' — 1 in
in as in i ^J- cn
m os in m
o o so CN
m, o m
as as in
so m SO
oo in in in o
os od in
in o cn as as
t-> *n cn i— * o
CN cN CN CN CN
CN CN CN CN CN
CN CN CN CN CN
CN CN CN CN CN
oo in ^ cn >— |
CN CN CN CN CN
<«t OS CN O O
oo in m cN
CN CN CN CN CN
as as oo oo
■sf m ■— i o
OO SO m CN
CN CN CN
oosO-^-mcN -h-hoo
cn cn cn cn cn cn cn cn cn
or~-ooooo ^HinOincN
assqinrncN cn-h-hOO
cncncncn'cn cn cn cn cn cn
CN CN CN CN CN
■^f O CN CN
as t — in ^j" m
CN CN CN CN CN
so cN so -^f "n
CN CN CN CN CN
oo in as
CN CN CN CN CN
— h oo CN < — 1 < — 1
p r-; sq in
en CN CN CN CN
CN CN CN CN
(T) CN CN CN CN
(T) CN CN CN CN
cn m cN CN CN
SO SO "-^ ' — 1 CN
^ cn — ; p as
rn rn rn rn cN
en en en en en
cn r~~ cn r-~ ^- <-<
CN —h h O p O
CN CN CN CN CN CN
CN CN CN CN CN
O r- t-
CN so m so
in os so as
CN CN CN CN CN
CN CN CN CN CN
CN CN CN
CN CN CN CN CN
^t-oomasso moooso^j-
cn cJ w --J h _H_Hppp
CNCNCNCNCN CN CN CN CN CN
Os Os OS OS OS
cN < — 1 os oo
O O OS OS OS
N (N H H* H
CN CN CN CN CN
in as in -h
en en en cn
CN CN CN CN CN
■^t- oo cn os so
in ^ m m
CN CN CN CN CN
CN CN CN CN CN
CN CN CN CN CN
CNCNCNCNCN -h,— i
CN CN CN CN CN
CN CN CN CN CN
SO *tf CN O OO
X ^ ^ ^1
CN CN CN CN CN
tJ- m ■— i o as
r-t r-i y-t y-t O
CN CN CN CN CN
CNCNCNCNCN CNCNCNCNCN
cn cn cn cn m
p p p as as
m rn cn cN CN
CN CN CN CN CN
CN CN CN CN CN
r-~ so in m
in in in in in
CN CN CN CN CN
so in m cN
OS OS OS OS OS
(N (N CN CN oi
CN CN ' — 1 " —|
CN CN CN CN
CN CN CN CN
OS ^ ^ so
""t cn
CN CN CN CN
CN CN CN CN
r—t cn cn tn so r-- oo as o -h cn en in so oo as o
Degrees of Freedom for the Denominator
Statistical Tables on the Web Site
Note: The following tables are on the Web site of the text along with Chapters 14 and 15.
Table VIII Critical Values of X for the Sign Test
Table IX Critical Values of T for the Wilcoxon Signed-Rank Test
Table X Critical Values of T for the Wilcoxon Rank Sum Test
Table XI Critical Values for the Spearman Rho Rank Correlation Coefficient Test
Table XII Critical Values for a Two-Tailed Runs Test with a = .05
ANSWERS TO SELECTED ODD-NUMBERED
EXERCISES AND SELF-REVIEW TESTS
(Note: Due to differences in rounding, the answers obtained by readers may differ slightly
from the ones given in this Appendix.)
Chapter 1
1.7
1.11
1.15
1.17
1.21
1.23
1.25
1.27
1.29
1.33
1.35
1.37
1.39
population b. sample c. population
sample e. population
number of dog bites reported last year
six observations c. six elements
quantitative b. quantitative c. qualitative
quantitative e. quantitative
discrete b. continuous d. discrete
continuous
cross-section data b. cross-section data
time-series data d. time-series data
1363
Xy = 45 c.
(ty) 2 = 2025
(2x) 2 = 732,736
(2x) 2 = 21,904
c. 2m/
txy
922
222
2/ = 69 b.
2m 2 / = 17,128
2x = 112 b.
2y 2 = 285 e.
2x = 856 b.
2x 2 = 157,574
2x = 148 b.
2x 2 = 4486
sample b. population for the year
sample d. population
sampling without replacement b. sampling with
replacement
a. 2x = 47 b. (2x) 2 = 2209 c. 2x 2 = 443
a. 2m = 59 b. 2/ 2 = 2662 c. 2m/ = 1508
d. 2m 2 / = 24,884 e. 2«
867
Self- Review Test
b 2. c 3. a. sampling without replacement
b. sampling with replacement
a. qualitative b. quantitative (continuous)
c. quantitative (discrete) d. qualitative
a. 2x = 29 b. (2x) 2 = 841 c. 2x 2 = 231
a. 2m = 45 b. 2/= 112 c. 2m 2 = 495
d. 2m/ = 975 e. 2m 2 / = 9855 f. 2/ 2 = 2994
Chapter 2
2.3 c. 26.7% d. 73.4%
2.5 c. 52% 2.7 c. 50% 2.15 d. 62%
2.17 a. class limits: $l-$25, S26-S50, $51-$75, $76-$100,
$101-$125, $126-$150 b. class boundaries:
$.5, $25.5, S50.5, $75.5, $100.5, $125.5, $150.5;
width = $25 c. class midpoints: $13, $38, $63, $8
$113, $138
2.19
2.35
2.47
2.67
2.73
2.87
d. 30% 2.29 e. 11
c. 38% e. about 52% 2.43 6 teams
218, 245, 256, 329, 367, 383, 397, 404, 427, 433, 471, 523,
537, 551, 563, 581, 592, 622, 636, 647, 655, 678, 689, 810,
841
d. 27.5% 2.69 c. 16.7% 2.71 c. 56.7%
d. Boundaries of the fourth class are $4200.5 and $5600.5;
width = $1400.
No. The older group may drive more miles per week than
the younger group.
Self-Review Test
2. a. 5 b. 7 c. 17 d. 6.5 e. 13
f. 90 g. .30
4. c. 35% 5. c. 70.8%
8. 30, 33, 37, 42, 44, 46, 47, 49, 51, 53, 53, 56, 60, 67,
67, 71, 79
Chapter 3
3.5
3.11
3.13
3.15
3.17
3.19
3.21
3.23
3.25
3.29
3.31
3.33
3.41
3.43
3.45
mode 3.9 mean = 3.00; median = 3.50; no mode
mean = $3779.44; median = $3250
a. mean = $272.98 billion; median = $162 billion
b. mode = $34 billion and $216 billion
mean = £48.515 million; median = £37.6 million
mean = $529.67 million;
median = $449.5 million; no mode
mean = 2.92 power outages; median = 2.5 power
outages; mode = 2 power outages
mean = 29.4; median = 28.5; mode = 23
a. mean = 1803; median = 1270 b. outlier = 5490;
when the outlier is dropped: mean = 1467.8;
median = 1 166; mean changes by a larger amount
c. median
combined mean = $148.89 3.27 total = $1055
age of the sixth person = 48 years
mean for data set I = 24.60; mean for data set II = 31.60
The mean of the second data set is equal to the mean of
the first data set plus 7.
10% trimmed mean = 38.25 years 3.35 weighted mean
= 77.5
range = 25; a 2 = 61.5; a = 7.84
a. x = 9; deviations from the mean: —2, 1, —1, —6, 6, 3,
— 3, 2. The sum of these deviations is zero.
b. range = 12; s 2 = 14.2857; s = 3.78
range = 13; s 2 = 13.8409; s = 3.72
ANl
AN2 Answers to Selected Odd-Numbered Exercises and Self-Review Tests
3.47
3.49
3.51
3.53
3.55
3.57
3.59
3.63
3.65
3.67
3.69
3.71
3.75
3.77
3.79
3.81
3.83
3.85
3.91
3.93
3.95
3.97
3.99
3.109
3.111
3.113
3.115
3.117
3.119
3.121
3.123
3.125
3.127
3.129
3.131
3.133
3.137
3.139
range = 27 pieces;
78.1;
7 stings;
„2
s 2 = 4.5769;
: 107.4286;
■ 151.7778;
j- = 8.84 pieces
s = 2.14 stings
= 10.36
range
range = 30;
range = 38;
s = 12.3198
s =
CV for salaries = 10.94%; CV for years of experience =
13.33%; The relative variation in salaries is lower.
i- = 14.64 for both data sets
5 = 9.40; s 2 = 37.7114; 5 = 6.14
H = 14 hours; a 2 = 51.9167; a = 7.21 hours
x = 19.67; s 2 = 67.6979; s = 8.23
x = 36.80 minutes; s 2 = 597.7143; s = 24.45 minutes
a. x = $139.05 c. x = $138.93
at least 75%; at least 84%; at least 89%
68%; 95%; 99.7%
at least 84%
at least 75%
at least 89%
i. at least 75%
$1515 to $3215
99.7% b. '
i. 99.7%
Qx = 69;
b.
ii. at least 89%
ii. 68%
& = 73; Q 3
c. 30.77%
Q 2 = 386.5; Q 3
c. 73.33%
Q 2 = 28.5; Q 3 =
c. 33.33%
Q 2 = 626.5; Q 3
b. P 30 = 552.5
c. 95%
b. 66 to 78 mph
= 76.5; IQR = 7.5
33; IQR = 8
= 728;
c. 23%
& = 369; Q 2 = 386.5; g 3 = 417; IQR = 48
P 5V = 390
Qx = 25;
^65 = 31
Qx = 533;
IQR = 195
no outlier
a. mean = $106.5 thousand; median = $76 thousand
b. outlier = 382; when the outlier is dropped: mean =
$75.9 thousand ; median = $74 thousand; mean changes
by a larger amount c. median
a. mean = 1973.6 points; median = 1917.5 points;
mode none b. range = 544 points;
s 2 = 42084.93; s = 205.15 points
x = 5.08 inches; s 2 = 6.8506; s = 2.62 inches
a. i. at least 75% ii. at least 89%
b. 160 to 240 minutes
a. i. 68% ii. 95% b. 140 to 260 minutes
a. g, = 60; Q 2 = 76; Q 3 = 97; IQR = 37
b. P 70 = 84 c. 70%
The data set is skewed slightly to the right; 135 is an outlier.
The minimum score is 169.
a. new mean = 76.4 inches; new median = 78 inches; new
range =13 inches b. new mean = 75.2 inches
mean = $54.46 per barrel
a. trimmed mean = 9.5 b. 14.3%
a. age 30 and under: rate for A = 25; rate for
B = 20 b. age 31 and over: rate for A = 100; rate
for B = 85.7 c. overall: rate for A = 50; rate for
B = 58.3 d. Country A has the lower overall average
because 66.67% of its population is under 30.
a. k = 1.41 b. k = 2.24 3.135 b. median
b. For men: mean = 82, median = 79, modes = 75, 79,
and 92, s = 12.08, Q x = 73.5, Q 3 = 89.5, and IQR = 16.
For women: mean = 97.53, median = 98, modes = 94
and 100, s = 8.44, Q x = 94, Q 3 = 101, and IQR = 7
a. mean = 30 b. mean = 50
3.141 a. at least 55.56% b. 1 to 11 inches
c. 2.66 to 9.34 inches
3.143 a. For men: mean = 174.91 lbs = 76,189.05 grams =
12.49 stone, median = 179 lbs = 77,970.61 grams =
12.79 stone, and st. dev. = 19.12 lbs = 8328.48 grams =
1.37 stone. For women: mean = 124.95 lbs = 54,426.97
grams = 8.93 stone, median = 123 lbs = 53,577.57 grams
= 8.79 stone, st. dev. = 17.48 lbs = 7614.11 grams =
1.25 stone. b. see answer to a, as answers are identical,
e. yes d & e. Smaller unit has more variability.
3.145 108 to 111
Self-Review Test
1. b 2. a and d 3.
6. b 7. a 8. a
12. c 13. a 14. a
15. mean = 10.9
c
9.
4.
c
10.
5. b
11. b
8; mode = 6; range = 26;
19.
20.
21.
22.
23.
24.
25.
26.
27.
= 65.2111;
x = 19.46;
i. at least 84%
2.9 to 1 1.7 years
i. 68% ii. 99.7%
Qx = 3; Q 2 = 8; Q }
median =
s = 8.08
s 2 = 44.0400; s = 6.64
ii. at least 89%
b.
= 13;
2.9 to 11.7 years
IQR = 10
b.
a.
b.
a.
a.
b.
Data are skewed slightly to the right,
combined mean = $1066.43
GPA of fifth student = 3.17
10% trimmed mean = 376.625; trimmed mean is a better
measure
a. mean for data set I = 19.75; mean for data set II = 16.75.
The mean of the second data set is equal to the mean of the
first data set minus 3. b. s = 11.32 for both data sets.
Chapter 4
4.3
4.5
4.7
4.9
4.11
4.13
4.19
4.21
4.23
4.25
4.31
4.35
4.39
4.47
4.49
S = {AB, AC, BA, BC, CA, CB)
four possible outcomes; S = {NN, NI, IN, II)
four possible outcomes; 5 = {DD, DG, GD, GG)
S = {HHH, HHT, HTH, HTT, THH, THT, TTH, TTT}
a. {NI and IN); a compound event
b. {II, NI, and IN}; a compound event
c. {NN, IN, and NI}; a compound event
d. {IN}; a simple event
a. {DG, GD, and GG); a compound event
b. {DG and GD); a compound event
c. {GD}; a simple event
d. {DD, DG, and GD); a compound event
-.55, 1.56, 5/3, -2/7
not equally likely events; use relative frequency approach
subjective probability
a. .450 b. .550 4.27 .660 4.29 .160
a. .200 b. .800 4.33 .6667; .3333
.325; .675 4.37 a. .0939 b. .5
use relative frequency approach 4.45 1296
a. no b. no c. A = {1, 3, 4^6, 8};
B = {1, 3, 5, 6, 7}; P(A) = .625; P(B) = .625
50 4.51 960 4.53 a. i. .600 ii. .600
iii. .375 iv. .583 b. Events "male" and
"female" are mutually exclusive. Events "have shopped"
and "male" are not mutually exclusive. c. Events
"female" and "have shopped" are dependent.
Answers to Selected Odd-Numbered Exercises and Self-Review Tests AN3
4.55
4.57
4.59
4.61
4.63
4.71
4.73
4.75
4.77
4.81
4.83
4.85
4.87
4.93
4.97
4.105
4.109
4.111
4.113
4.115
4.119
4.125
4.127
4.129
4.131
4.135
4.139
4.141
a. i. .3475 ii. .5425 iii. .2727
iv. .4545 b. Events "male" and "in favor" are not
mutually exclusive. Events "in favor" and "against" are
mutually exclusive. c. Events "female" and "no
opinion" are dependent.
a. i. .1012 ii. .4835 iii. .5524
iv. .1014 b. Events "Airline A" and "more than 1
hour late" are not mutually exclusive. Events "less than 30
minutes late" and "more than one hour late" are mutually
exclusive. c. Events "Airline B" and "30 minutes to
1 hour late" are dependent.
Events "female" and "pediatrician" are dependent but not
mutually exclusive.
Events "female" and "business major" are dependent but
not mutually exclusive.
P(A) = .3333; P(A) = .6667 4.65 .88
a. .4543 b. .0980
a. .1520 b. .1824
a. .2462
.6923
a. i.
a. i.
a. i.
.3529
b. .0000
.0000
.1600
.5120
.52
b. .67
.1086
.725
ii. .1590
ii. .150
ii. .035 b.
.2667 4.91
.9025 4.95
.40
76 4.107 a.
.9075
.750 c. 1.0
.550 c. .790
77
80 4.123 .9744
.1429
ii. .4800 iii. .3462
.3400 vi. .6600
b. Events "female" and "prefers watching sports" are
dependent but not mutually exclusive.
a. i. .750 ii. .700 iii. .225 iv. .775
b. Events "student athlete" and "should be paid" are
dependent but not mutually exclusive.
a. .7242 b. .2758 4.133 .0605
.0048 4.137 a. 17,576,000 b. 5200
a. 1/195,249,054 = .0000000051
b. 1/5,138,133 = .00000019
a. .5000 b. .3333 c. No; the sixth toss
is independent of the first five tosses. Equivalent to part a.
4.79
.3844
.350
.225
4.89
a. .0025 b.
.5278 4.99
a. .56 b. .
a. .6358 b.
a. .750 b.
a. .780 b.
.910 4.117 .
.700 4.121 .
a. .2571 b.
a. i. .4360
iv. .6809 v.
4.143
a.
.030
b. .150
4.145
a.
.50
b. .50 4.147
a. .8333
b. .1667
4.149
a.
.01%
b. i. .0048
ii. .0028
iii.
.0222
iv. .0111
4.151 a.
.8851
b.
.0035
Self-Review Test
1. a 2. b 3. c 4. a 5. a 6. b
7. c 8. b 9. b 10. c 11. b
12. 120 13. a. .3333 b. .6667
14. a. Events "female" and "out of state" are dependent but not
mutually exclusive. b. i. .4500 ii. .6364
15. .825 16. .3894 17. .4225 18. .40; .60
20. a. i. .358 ii.
b. Events "woman"
mutually exclusive.
.405 iii. .235 iv. .5593
and "yes" are dependent but not
Chapter 5
5.3
5.5
5.9
5.25
5.27
5.29
5.31
5.33
5.35
5.37
5.39
5.41
5.45
5.47
5.51
a. discrete random variable b. continuous random
variable c. continuous random variable
d. discrete random variable e. discrete random
variable f. continuous random variable
discrete random variable
a. not a valid probability distribution b. a valid
probability distribution
distribution
c. not a valid probability
5.11
a.
.17 b.
.20
c.
.58
d.
.42
e.
.42 f.
.27
g-
.68
5.13
b.
i. .51
ii.
.235
iii.
.285
iv.
.305
5.15
a.
X
1
2
3
4
5
P(x)
.10
.25
.30
.20
.15
b.
iii.
approximate
.75 iv.
.30
.65
.65
5.17
X
1
2
P(x)
.5271
.3978
.0751
5.19
X
1
2
P(.x)
.3969
.4662
.1369
5.21
X
1
2
P(x)
.4789
.4422
.0789
5.23
a. jju
= 1.590;
cr = .960
b.
a = 1.061
V =
M =
^ =
M =
M =
M =
M =
3! =
(14
4C0
.440 error; cr =
2.94 camcorders;
1.00 head; a =
2.5604 tires;
.100 lemon;
$3.9 million;
.500 person;
6; (9 - 3)!
- 12)! = 2;
H = 7.070;
852 error
cr = 1.441 camcorders
707 head
cr = 1.3223 tires
cr = .308 lemon
cr = $3,015 million
cr = .584 person
= 720; 9! = 362,880;
5 C 3 = 10; 7 C 4 = 35; 9 C 3 = 84;
3 C 3
1;
= 72
20-^6 =
j = 30
5.43 12 <
27,907,200
1;
9 C 2 = 36; 9 J
20 C 6 = 38,760
167,960
a. not a binomial experiment
b. a binomial experiment
c. a binomial experiment
^4
1680
220; ,
,P 3 = 1320
19. a. .279
b. .829
5.53
a.
.2541
b.
.1536
c.
.3241
5.55
b.
/x = 2.100;
cr = 1.212
5.59
a.
0, 1, 2, 3,
4.
5, 6, 7, 8,
9, 10, 11 b. .
1161
5.61
a.
.0314
b.
.3552
c.
.8076
5.63
a.
.0913
b.
.0000
c.
.0122
5.65
a.
.2725
b.
.0839
5.67
a.
/i = 5.6 customers; cr
= 1.058 customers
b.
5.69
a.
\x, = 5.600 customers;
cr =
1.296 customers
b.
.0467
5.71
a.
.4286
b.
.0714
c.
.5
5.73
a.
.3818
b.
.0030
c.
.5303
5.75
a.
.4747;
b.
.0440
c.
.3407
5.77
a.
.1078
b.
.5147
c.
.8628
5.81
a.
.0404
b.
.2565
5.83
a.
ix, = 1.3; a 2
= 1.3; 0- :
= 1.140 b. ix.
= 2.1:
cr 1
= 2.1; cr --
= 1.449
.1147
AN4 Answers to Selected Odd-Numbered Exercises and Self-Review Tests
5.85
.1496 5.87
.1185
6.41
a.
93.32%
b.
15.57%
5.89
a. .1162
b.
.6625
ii.
.1699
6.43
a.
.0197
b.
.3296
iii. .4941
6.45
a.
.8264
b.
12.83%
5.91
a. .3033
b.
.0900
ii.
.0018
6.47
a.
15.62%
b.
7.64%
iii. .9098
6.49
a.
0.39%
b.
1.46% c. 18.72%
5.93
a. .0031
b.
.0039
ii.
.4911
d.
29.21%
5.95
a. .2466
c.
M =
1.4 a 2
= 1.4
6.51
2.64%
a = 1.183
6.53
a.
2.00
b. -
-2.02 approximately
5.97
5.99
5.101
5.103
5.105
5.107
5.109
5.111
5.113
5.115
5.117
5.119
5.121
5.123
5.129
.0390
.2580
a. .0446 b.
iii. .0218
ft = 4.11; cr = 1.019; This mechanic repairs, on average,
4.1 1 cars per day
b. /jl = $557,000; cr = $1,288,274; ^ gives the
company's expected profit.
a. .0000 b. .0351 c. .7214
a. .9246 b. .0754
a. .3692 b. .1429 c. .0923
a. .8643 b. .1357
a. .0912 b. i. .5502 ii. .0817
iii. .2933
a. .2466
Sx P(x) = —2.22. This game is not fair to you and you
should not play as you expect to lose $2.22.
a. .0625 b. .125 c. .3125
c. .7149 d. 3 nights
8 cheesecakes
a. 35 b. 10 c. .2857 5.127 $6
a. .0211 b. .0475 c. .4226
Self-Review Test
8.
2. probability distribution table
3. a 4. b 5. a 7. b
9. b 10. a 11. c 13. a
15. jx = 2.040 homes; a = 1.449 homes
16. a. i. .2128 ii. .8418 iii. .0153
b. jli = 7.2 adults; a = 1.697 adults
17. a. .4525 b. .0646 c. .0666
18. a. i. .0521 ii. .2203 iii. .2013
Chapter 6
6.11
.8664 6.13 .9876
6.15
a.
.4744 b. .4798
c.
.1162
d. .0610
e.
.9400
6.17
a.
.0869 b. .0244
c.
.9798
d. .9608
6.19
a.
.5 approximately
b. .5 approximately
c.
.00 approximately
d.
.00 approximately
6.21
a.
.9613 b. .4783
c.
.4767
d. .0694
6.23
a.
.0096 b. .2466
c.
.1570
d. .9625
6.25
a.
.8365 b. .8947
c.
approximately .5
d.
approximately .5
e. approximately .00
f.
approximately .00
6.27
a.
1.80 b. -2.20
c.
-1.20
d. 2.80
6.29
a.
.4599 b. .1598
c.
.2223
6.31
a.
.3336 b. .9564
c.
.9686
d.
approximately .00
6.33
a.
.2178 b. .6440
6.35
a.
.8212 b. .2810
c.
.0401
d. .7190
6.37
a.
.0764 b. .1126
6.39
a.
.0985 b. .0538
6.55
6.57
6.59
6.61
6.63
6.65
6.67
6.69
6.71
6.73
6.77
6.79
6.81
6.83
6.85
6.87
6.89
6.93
6.95
6.97
6.99
6.101
6.105
c. —.37 approximately d. 1.02 approximately
a. approximately 1.65 b. —1.96 c. —2.33
approximately d. 2.58 approximately
a. 208.50 b. 241.25 c. 178.50
d. 145.75 e. 158.25 f. 251.25
19 minutes approximately
2060 kilowatt-hours
$82.02 approximately
np > 5 and nq > 5
a. .7688 b. .7697; difference is
.0009
.3192
a. fjL = 72; cr = 5.36656315 b.
e. .4564
a. .0764 b. .6793 c. .8413 d. .8238
.0735 6.75 a. .0351 b. .1875 c. .1230
a. .0454 b. .0516 c. .8646
a. .7549 b. .2451
a. .1093 b. 9.31% c. 57.33%
d. It is possible, but its probability is close to zero.
.0124 or 1.24%
a. 848 hours b. 792 hours approximately
a. .0454 b. .0838 c. .8861 d. .2477
$2136 6.91 a. 85.08% b. $4000
.0091
a. at most .0062 b. 65 mph
8.16 ounces
company A: $.0490 company B: $.0508
a. .7967 b. 62
.1064
Self-Review Test
1. a 2. a 3. d 4. b 5. a 6. c
7. b 8. b
9. a. .1878 b. .9304 c. .0985 d. .7704
10. a. -1.28 approximately b. .61 c. 1.65
approximately d. —1.07 approximately
11. a. .5608 b. .0015 c. .0170 d. .1165
12. a. 48669.8 b. 40162
13. a. i. .0318 ii. .9453 iii. .9099
iv. .0268 v. .4632 b. .7054 c. .3986
Chapter 7
7.5
7.7
7.13
7.15
a.
c.
d.
b.
a.
b.
a.
16.60 b. sampling error = —.27
sampling error = —.27; nonsampling error = 1.11
Xi = 16.22; x 2 = 15.67; x 3 = 17.00; x 4 = 16.33;
x b = 16.78; x 7 = 17.22;
Xi, = 16.56; x U) = 15.11
x s = 17.44;
x s = 17.67;
3cj = 28.4;
x 5 = 35.2;
= 60;
/Us = 60;
cr T = 1.400
l 10
c 3 = 33.8;
c. fl =
x 2 = 28.8;
x b = 36.4;
a x = 2.357
cr s = 1.054
b. a- r = 2.500
34.4;
32.83
Answers to Selected Odd-Numbered Exercises and Self-Review Tests AN5
7.17 a. n = 100 b. n = 256
7.19 /xj = S3.084; <r x = $.038
7.21 il- x = $520; <t x = $14.40 7.23 n = 256
7.25 a. fi s = 80.60 b. a x = 3.302
d. oj = 3.302
7.33 /u^ = 20.20 hours; cr^ = .613 hours; the normal
distribution
7.35 /if = 3.020; Oj = .042; approximately normal
distribution
7.37 n = 20: fi x = 91.4 grams; <j- x = 20.851 grams; skewed to
the right
n = 75: fi x = 91.4 grams; cr- x = $10,768 grams;
approximately normal distribution
7.39 /u^ = 200 pieces; cr T = 15.821 pieces; approximately
normal distribution; no, sample size a 30
7.41 86.64%
7.43
a.
z = 2.44
b. z =
-7.25
c.
z = -3.65
d.
z = 5.82
7.45
a.
.1940
b.
.8749
7.47
a.
.0003
b.
.9292
7.49
a.
.1093
b.
.0322
c.
.7776
7.51
a.
.0150
b.
.0968
c.
.5696
7.53
a.
.8203
b.
.9750
7.55
a.
.1147
b.
.9164
c.
.1251
7.57
a.
.1032
b.
.3172
c.
.0016
d. .9049
7.59
.0124 7.61
P = -12;
p =
15
7.63 7125 subjects in the population; 312 subjects in the
sample
7.65 sampling error = —.05
7.71 a. n t = .21; crp = .020
b. up = .21; crp = .015
7.73 a. crp = .051 b. crp = .071
7.77 a. p = .667 b. 6 d. -.067, -.067, .133, .133,
-.067, -.067
7.79 /jip = .30; crp = .034; approximately normal distribution
7.81 fip = .561; crp = .027; approximately normal
distribution
7.83 95.44%
7.85 a. z = -.61 b. z = 1.83 c. z = -1.22
d. z = 1.22
7.87 a. .0721 b. .1798
7.89 a. .0030 b. .2678
7.91 .2005
7.93 /u,j = 750 hours; cr x = 11 hours; the normal distribution
7.95 a. .9131 b. .1698 c. .8262 d. .0344
7.97 a. .489 b. .0006 c. .8064 d. .8643
7.99 = .88; crp = .036; approximately normal distribution
7.101 a. i. .0146 ii. .0907 b. .9912 c. .0146
7.103 .6318
7.105 10 approximately
7.107 a. .8023 b. 754 approximately
7.109 .0035
Self-Review Test
1. b 2. b 3. a 4. a 5. b
6. b 7. c 8. a 9. a
10. a 11. c 12. a
14. a. pjj, = 145 pounds; cr T = 3.600 pounds; approximately
normal distribution
b. yiij = 145 pounds; cr 5 = 1.800 pounds; approximately
normal distribution
15. a. /Aj = 45,000 miles; = 527.71 miles; unknown
distribution
b. /Xj = 45,000 miles; <x T = 292.72 miles; approximately
normal distribution
16. a. .1541 b. .4582 c. .0003 d. .1706
e. .0084
17. a. i. .1203 ii. .1335 iii. .7486
b. .9736 c. .0013
18. a. /Jip, = .048; crp = .0302; unknown distribution
b. fip = .048; crp = .0096; approximately normal
distribution
c. /Jip = .048; crp = .0030; approximately normal
distribution
19. a. i. .0080 ii. .4466 iii. .7823
iv. .2815 b. .5820 c. .1936 d. .0606
Chapter 8
8.11
a.
24.5 b. 22.71
to 26.29 c. ±1.79
8.13
a.
70.59 to 79.01
b. 69.80 to 79.80
c.
68.22 to 81.38
d. yes
8.15
a.
77.84 to 85.96
b. 78.27 to 85.53
c.
78.65 to 85.15
d. yes
8.17
a.
38.34 b. 37.30 to 39.38 c. ±1.04
8.19
a.
n = 167 b. n
= 65
8.21
a.
n = 299 b. n
= 126 c. n = 61
8.23 $295,146.86 to $304,293.14
8.25 a. 48,903.27 to 58,196.73 labor-hours
8.27 31.86 to 32.02 ounces; no adjustment needed
8.29 a. $1532.41 to $1617.59
8.31 n = 167 8.33 n = 61
8.41 a. t= -1.325 b. t =2.160 c. t =3.281
d. t = -2.715
8.43 a. a « .10, left tail b. a = .005, right tail
c. a = .10, right tail d. a ~ .01 left tail
8.45 a. t = 2.080 b. t = 1.671 c. t= 2.807
8.47 a. 1.41 b. -3.40 to 6.22 c. ±4.81
8.49 a. 24.06 to 26.94 b. 23.58 to 27.42
c. 23.73 to 27.27
8.51 a. 91.03 to 93.87 b. 90.06 to 93.44
c. 88.06 to 91.20 d. confidence intervals of parts b
and c cover /x, that of part a does not
8.53 40.04 to 42.36 bushels
8.55 .32 to .36 grams
8.57 18.64 to 25.36 minutes
8.59 a. 21.56 to 24.44 hours
8.61 4.88 to 11.12 hours
8.63 7.20 to 8.14 ounces
8.65 a. 6.18 years
b. 5.85 to 6.51 years; margin of error: ±.33 year
8.71 a. yes, sample size is large b. no, sample size is
not large c. yes, sample size is large d. yes,
sample size is large
8.73 a. .297 to .343 b. .336 to .384
c. .277 to .323 d. confidence intervals of parts a and
b cover p, but that of part c does not
8.75 a. .189 to .351 b. .202 to .338
c. .218 to .322 d. yes
AN6 Answers to Selected Odd-Numbered Exercises and Self-Review Tests
8.77 a. .284 to .336 b. .269 to .351
c. .209 to .411 d. yes
8.79 a. n = 668 b. n = 671
8.81 a. n = 1432 b. n = 196 c. n = 353
8.83 a. .29 to .45
8.85 a. 40% b. 33.1% to 46.9%; margin of error = ±6.9%
8.87 a. 20.3% to 55.7% 8.89 a. 12.3% to 17.1%
8.91 a. .333 b. 8.5% to 58.1%
8.93 n = 1084
8.95 n = 1849
8.99 a. $2640
b. $2514.57 to $2765.43
8.101 3.969 to 4.011 inches; the machine needs to be adjusted
8.103 12.5 to 16.5 gallons
8.105 21.76 to 26.24 minutes
8.107 4.4 to 4.6 hours
8.109 144.33 to 158.47 calories
8.111 a. .03 b. .014 to .046
8.113 6.1% to 56.4%
8.115 n = 221 8.117 n = 359
8.121 n = 65
8.123 a. h — 20 days b. 90% c. ±75 cars
Self- Review Test
1. a. population parameter; sample statistic
b. sample statistic; population parameter
c. sample statistic; population parameter
2. b 3. a 4. a 5. c 6. b
7. a. $159,000
b. $147,390 to $170,610; margin of error = ± $1 1,610
8. $379,539.30 to $441,310.70 9. a. .55
b. .489 to .611
10. n = 83 11. * = 273 12. n = 229
Chapter 9
9.5 a. a left-tailed test b. a right-tailed test
c. a two-tailed test
9.7 a. Type II error b. Type I error
9.9 a. H : p = 20 hours; H t : p j= 20 hours;
a two-tailed test b. H : /x = 10 hours;
H{. fi > 10 hours; a right-tailed test
c. H : p = 3 years; Hy p + 3 years; a two-tailed test
d. H : /x = $1000; H t : fx < $1000;
a left-tailed test e. H : fx — 12 minutes;
Hi. /x > 12 minutes; a right- tailed test
9.17 a. p-value = .0188 b. p-value = .0116
c. p-value = .0087
9.19 a. p-value = .0166 b. no, do not reject H
c. yes, reject H a
9.21 a. rejection region is to the left of —2.58 and to the right
of 2.58; nonrejection region is between —2.58 and 2.58
b. rejection region is to the left of —2.58; nonrejection
region is to the right of —2.58
c. rejection region is to the right of 1.96; nonrejection
region is to the left of 1.96
9.23 Statistically not significant
9.25 a. .10 b. .02 c. .005
9.27 a. observed value of z is .58; critical value of z is ±1.96
b. observed value of z is .58; critical value of z is 1.65
9.29 a. reject H if z > 1.65 b. reject H if z < -1.65
c. reject H if z < ~ 1 .96 or z > 1 .96
9.31 a. critical value: z = —1.96; test statistic: z = —2.67;
reject H b. critical value: z = —1.96; test statistic:
z = —1.00; do not reject H
9.33 a. critical values: z = —1.65 and 1.65; test statistic: z =
— 1.34; do not reject H u b. critical value: z = —2.33;
test statistic: z = —6.44; reject H Q
c. critical value: z = 1.65; test statistic: z = 8.70;
reject H
9.35 a. H : p = 45; ff t : fx < 45 months; p value = .0170; if
a = .025 reject H b. test statistic:
z = —2.12; Critical value: z = —1.96; reject H a
9.37 a. H : fx > 25 years; H, < 25 years; p value = .0418; if
a = .025, do not reject H
b. Critical value: z = —1.96; observed value: z = —1.73;
do not reject H
9.39 a. H Q : jx =10 minutes; H { : /x ^10 minutes; test statistic:
z = -2.11; p value = .0348 If a = .02, do not reject H . If
a = .05, reject H . b. Observed value z = —2.11;
If a = .02, critical values: z = —2.33 and 2.33; do not
reject H a . If a = .05, critical values: z = —1.96 and 1.96;
reject H Q .
9.41 a. test statistic: z = -2.33; p value = .0198;
If a = .01; do not reject H ; If a = .05, reject H .
b. Observed value z = —2.33; If a = .01, critical values:
z = —2.58 and 2.58, do not reject H : If a = .05, critical
values: z = —1.96 and 1.96; reject H .
9.43 a. H : fx > 47.93 boxes; H{. fx < 47.93 boxes; critical
value: z = —1.28; test statistic: z = —1.16; do not reject H
b. reject H .
9.45 a. H : fi > 8 hours; H\. fx < 8 hours; critical value: z =
-2.33; a = .01; test statistic: z = -.68; p value = .2483
do not reject H . b. critical value: z = —1.96;
test statistic: z = —.68; do not reject H .
9.49 a. reject H u if t < -2.977 or t > 2.977 b. reject H
if t < -2.797 c. reject H if t > 2.080
9.51 a. critical values: t = —2.365 and 2.365; observed value:
t = -2.097; .05 < p value < .10 b. critical value:
t = -1.895; observed value: t = -2.097; .025 < p value
< .05
9.53 a. reject H if t> 1.672 b. reject H Q if t <- 1.672
c. reject H if t < -2.002 or t > 2.002
9.55 a. critical value: t = 1.998; test statistic: t = 4.800;
reject H b. critical value: t = 1.998; test statistic:
/ = 1.143; do not reject H
9.57 a. critical value: ; = -1.363; test statistic: t =-1.252;
do not reject H b. critical values: t = —2.064 and
2.064; test statistic: / = 2.258; reject H c. critical
value: t = 3.143; test statistic: t = 2.658; do not
reject H
9.59 H : fx ^ 4.145 minutes; //,: fx > 4.145 minutes;
critical value: t = 1.301; test statistic: t = 1.862;
reject H ; .025 < p value < .05; reject H
9.61 H : /x = $850; H x : fx < $850; critical value: t = -2.397;
test statistic: t = —2.257; do not reject H ; if a = .025,
critical value = —2.005; reject H Q
9.63 H : fx = 14.325 homes; H x : fi + 14.325 homes; test
statistic: t = —.752; p-value > .10; do not reject H ;
for a = .05, critical values: t = -2.020 and t = 2.020;
test statistic: t = —.752; do not reject H
Answers to Selected Odd-Numbered Exercises and Self-Review Tests AN7
9.65 a. H : p s= $150; Hi: /x < $150; test statistic: t = -1.964;
.025 < p-value < .050; do not reject H (l ; for a = .01,
critical value: t = -2.492; test statistic: t = -1.964; do
not reject H b. a = .01
9.67 a. H Q : p = 58 years; H x : p + 58 years; if a = 0, do not
reject H b. test statistic: t = —4.183; p- value <
.002; for a = .01, reject H ; critical values: t = -2.649
and 2.649; test statistic: t = -4.183; reject H
9.69 H : /x = $95; H,: /x > $95; critical value: t = 1.771; test
statistic: t = 2.130; reject H
9.71 H : /x = $3,084; H ( : /x < $3,084; test statistic:
t = -2.024; .01 < p value < .025; do not reject H ;
critical value: t = —2.326; do not reject H
9.75 a. not large enough b. large enough
c. not large enough d. large enough
9.77 a. reject H if z < -1.65 or z > 1.65 b. reject H if
z < -2.33 c. reject H„ if z > 1.65
9.79 a. critical value: z = 1.65; observed value: z = 3.90
b. critical values: z = —1.96 and 1.96; observed value:
z = 3.90
9.81 a. reject H if z < - 1.65 b. reject H if z < - 1.96
or z > 1.96 c. reject H if z > 1.65
9.83 a. critical values: z = —2.58 and 2.58; test statistic: z =
— 1.07; do not reject H b. critical values: z = —2.58
and 2.58; test statistic: z = 3.21; reject H ()
9.85 a. critical values: z = —1.65 and 1.65; test statistic:
z = .80; do not reject H b. critical value: z = —1.65;
test statistic: z = —4.71; reject H c. critical value:
z = 2.33; test statistic: z = .93; do not reject H
9.87 H : p = .30; H t : p < .30; test statistic: z =-1.75;
p value = .0401; for a = .05, reject H ; critical value for
a =.05: z = —1.65; test statistic: z = —1.75; reject H ()
9.89 H () : p = .67; H,: p > .67; critical value: z = 2.05; test
statistic: z = 3.15; reject H ; p-value =.0009; for a- = .02,
reject H
9.91 H () : p = .231; H, = p > .231; critical value: z = 1.65;
test statistic: z = .97; do not reject H ; p value = .1660;
for = .05, do not reject H
9.93 a. H„: p > .35; H,: p < .35; critical value: z = -1.96;
test statistic: z = —2.94; reject H b. do not reject H
c. a- = .025; p value = .0016; reject H
9.95 a. critical value: z = 1.96; test statistic: z = 2.27; reject H ;
adjust machine b. critical value: z = 2.33; test statistic:
z = 2.27; do not reject H ; do not adjust the machine
9.99 a. critical value: z = 1.96; test statistic: z = 2.10; reject
H () b. />(Type I error) = .025
c. p-value = .0179; do not reject H () if a = .01; reject H
if a = .05
9.101 a. critical values: z = —2.33 and 2.33; test statistic:
z = 2.55; reject H () b. />(Type I error) = .02
c. p-value = .0108; reject H if a = .025; do not reject
H () if a = .005
9.103 a. H : /x = 67.2 minutes; H^ p > 67.2 minutes; test
statistic: z = 4.98; p value < .0002; if a = .05, reject H
b. critical value: z = 2.33; test statistic: z = 4.98; reject
H
9.105 a. H„: /x > 50; Hj: /x < 50; critical value of z = - 1.96;
test statistic: z = —3.00; reject H
b. /"(Type I error) = .025 c. do not reject H
d. p-value = .0013; for a = .025, reject H
9.107 a. H : p =S 2400 square feet; H x : /x > 2400 square feet;
critical value: t = 1.677; test statistic: t = 2.097; reject H
b. for a = .01 critical value: / = 2.405; test statistic:
/ = 2.097; do not reject H
9.109 H : ft =£ 15 minutes; H,: /x > 15 minutes; critical value:
f = 2.438; test statistic: t = 1.875; do not reject H
9.111 H : p. = 25 minutes; /x # 25 minutes; critical values: ?
= -2.947 and 2.947; test statistic: t = 2.083; do not
reject H
9.113 a. H () : /x =£ 2 hours; H^ /x > 2 hours; critical value: f =
2.718; test statistic: r = 1.679; do not reject H„
9.115 a. H : p = .5; H^ p ^ .5; critical values: z = -1.96
and 1.96; test statistic: z = 1.52; do not reject H
b. /'(Type I error) = .05 c. a = .05; p-value =
.1286; do not reject H u .
9.117 H : p = .40; H ( : p =h .40; critical values: z = -2.58 and
2.58; test statistic: z = —1.62; do not reject H u ; p-value
= .1052; do not reject H
9.119 a. H : p = .80; H,: p < .80; critical value: z = -2.33;
test statistic: z = —.79; do not reject H ( ,
b. do not reject H
9.121 a. .0238 b. a = .0238
9.123 a = .3446
9.125 H : /jl = 750 hours; H,: /x < 750 hours; reject H
if x < 735: a = .0082; reject H if 5 < 700: a = .0000
9.129 a. 29 or more, or 1 1 or less b. 226 or more, or 174
or less c. 2081 or more, or 1919 or less
Self-Review Test
1. a 2. b 3. a 4. b 5. a 6. a
7. a 8. b 9. c 10. a 11. c 12. b
13. c 14. a 15. b
16. a. H : /x = 90.25 ozs; H,: /x # 90.25 ozs; critical values:
z = —2.58 and 2.58; test statistic: z = —3.18; reject H ()
b. H : /x = 90.25 ozs; H x : p, > 90.25 ozs; critical value:
z = —1.96; test statistic: z = —3.18; reject H c. in part a,
a = .01; in part b, a = .025 d. p-value = .0014, reject
H e. p-value = .0007, reject H
17. a. H : /x = 185; H^ /x < 185; critical value: t = -2.438
test statistic: t = —3.000; reject H b. /'(Type I error)
= .01 c. do not reject H d. .001 < p-value <
.005; for a = .01, reject H„
18. a. H : /x > 31 months; H,: /x < 31 months; critical value:
f = -2.131; test statistic: ? = -3.333; reject H
b. P(Type I error) = .025 c. critical value:
t = —3.733; do not reject H
19. a. H : p = .5; H,: p < .5; critical value: z = —1.65; test
statistic: z = —3.16; reject H b. /'(Type I error) = .05
e. do not reject H d. p-value = .0008; reject H () if
a = .05; reject H if a = .01
Chapter 10
10.3 a. 1.83; b. -.72 to 4.38; margin of error = ±2.55
10.5 H : p., — /x 2 = 0; H{. p x — p 2 ^ 0; critical values:
z = —1.96 and 1.96; test statistic: z = 1.85; do not reject H
10.7 H : /Xj — /x 2 = 0; H^ xq — /x 2 < 0; critical value:
z = —1.65; test statistic: z = —1.47; do not reject H
10.9 a. 9 hours b. 1.65 to 16.35 hours; c. H :
p l — /x 2 = 0; Hi: /X, — /x 2 0; critical values: z = —2.33
AN8
Answers to Selected Odd-Numbered Exercises and Self-Review Tests
and 2.33; test statistic: z = 2.66; reject H \ p-value = .0078;
for a = .02, reject H
10.11 a. .74 b. .373 to 1.11 c. H : p, x - p^ = 0;
H x : Hi — /x. 2 > 0; critical value: 2.33; test statistic: z =
3.95; reject H ; p value = .0000; for a =.01, reject //
10.13 a. -$1024.54 to -$75.46 b. H : p x - /x 2 = 0; //,:
p x — p 2 =/= 0; critical values: z = —2.58 and 2.58; test
statistic: z = —2.99; reject H Q c. do not reject H
10.15 a. -6.87 to .87 calories b. H : p x - ai 2 = 0; i^:
P-i ~ Pa< 0; critical value: z = —2.58; test statistic:
z = —1.81; do not reject H c. p value = .0351; do
not reject H for a = .005; do not reject H for a = .025
10.17 a. -1.58 b. -3.82 to .66
10.19 H : p x — p 2 = 0; H x : p x — p 2 + 0; critical values: t =
-2.023 and 2.023; test statistic t = - 1.430; do not reject H
10.21 H : p x — p 2 = 0; H x : p x — p 2 < 0; critical value:
t = -2.426; test statistic: t = -1.430; do not reject H
10.23 a. 2.62 b. -5.85 to 11.09 c. H : p x - p 2 = 0;
H x : p x — p 2 > 0; critical value: t = 2.500; test statistic:
t = .77; do not reject H
10.25 a. -46.80 to -7.20 miles; b. H : p x - p 2 = 0;
H x : p x — p 2 < 0; critical value: t = —2.326;
test statistic: t = —2.67; reject H
10.27 a. 2.29 to 5.71 mph b. H : p x - p^ = 0;
//]: p x — p 2 > 0; critical value: f = 2.416; test statistic:
f = 5.658; reject H
10.29 a. -12.95 to 2.95 minutes b. H : p^ - p 2 = 0;
pi — p 2 < 0; critical value: f = —2.412; test statistic:
f = — 1.691; do not reject ff
10.31 a. -.61 to -.39 b. H : p x - p 2 = 0; ffj:
l^i ~ f 1 ! ^ 0; critical values: f = —2.576 and 2.576; test
statistic: / = —10.130; reject H
10.33 -7.86 to -1.04
10.35 H : pi p 2 — 0; H^, p x — p 2 + 0; critical values:
t = -2.101 and 2.101; test statistic: t = -2.740; do not
reject H
10.37 H : p x — p 2 = 0; H{. p x — p 2 < 0; critical value:
t = -2.552; test statistic: t = -2.740; reject H
10.39 a. -47.01 to -6.99 miles; b. H : p { - p 2 = 0;
//]: pi — p 2 < 0; critical value: f = —2.326; test statistic:
t = -2.64; reject H c. -48.30 to -5.70; critical
value: t = —2.397; test statistic: t = —2.54; reject H
10.41 a. 2.23 to 5.77 mph b. H : p l - p 2 = 0; H t : p t -
p 2 > 0; critical value: t = 2.445; test statistic: t = 5.513;
reject H c. 1.81 to 6.20 mph; critical value:
t = 2.492; test statistic: t = 4.541; reject H
10.43 a. -12.86 to 2.86 minutes b. H : p, - p 2 = 0; H x :
AM — 1^2 < 0; critical value: ? = —2.414; test statistic:
f = -1.713; do not reject H c. -13.34to3.34
minutes; critical value: t = —2.431; test statistic:
t = — 1.63; do not reject H
10.45 a. -.61 to -.39 b. H : p x - p 2 = 0; H x :
Mi — l^i ^ 0; critical values: t = —2.576 and 2.576; test
statistic: r = -10.162; reject H c. -.62 to -.38;
critical values: t = —2.576 and 2.576; test statistic:
t = -10.10; reject H
10.49 a. 11.85 to 23.15 b. 50.08 to 61.72; c. 25.66
to 32.94
10.51 a. critical values: t = —2.060 and 2.060; test statistic:
t = 12.551; reject H b. critical value: t = 2.624;
test statistic: t = 7.252; reject H c. critical value:
t = -1.328; test statistic: t = -14.389; reject H
10.53 a. -2.98 to 9.84 minutes b. H : p d = 0; H x :
p d > 0; critical value: t = 2.447; test statistic: t = 1.983;
do not reject H Q
10.55 a. 13.22 to 30.01 seconds b. H : p d < + 15;
Hi. p d +15; critical value: t = 1.356; test statistic:
t = 1.72, reject H
10.57 a. -1.02 to 1.52 b. H : p d = 0; H t : p d + 0;
critical values: t = —2.093 and 2.093; test statistic:
t = .4122; do not reject H
10.61 -.062 to .142
10.63 H : Pi — p 2 = 0; H^. p x — p 2 # 0; critical values:
z = —1.96 and 1.96; test statistic: z = .76; do not
reject H u
10.65 Hq. Pi ~ p 2 = 0; H{. p x — p 2 > 0; critical value:
z = 2.05; test statistic: z = .76; do not reject H
10.67 a. -.04 b. -.086 to .006 c. rejection region
to the left of z = —2.33; non-rejection region to the right of
Z = -2.33 d. test statistic: z = -2.02 e. do not
reject H
10.69 a. - .025 to .225 b. H : p x - p 2 = 0; H x : p, - p 2
< 0; critical values: z = 1.96 and —1.96; test statistic:
z = 1.56; do not reject H a ; p value =.1188; for a =.025,
do not reject H c. .012 to .188; critical values: z =
— 1.96 and 1.96; test statistic: z = 2.20; reject H ; p value
= .0278; for a =.05, reject H
10.71 a. .024 b. -.020 to .068 c. H :p l -p 2 = Q\
Hi- Pi ~ V2 ^ 0; critical values: z = —1.96 and 1.96; test
statistic: z = 1.09; do not reject H ; p-value = .2758; for
a = .05, do not reject H
10.73 a. .10 b. .018 to .182 c. H : Pl - p 2 = 0;
H\- Pi ~ P2 ^ 0; critical values: z = —2.58 and 2.58; test
statistic: z = 3.04; reject H
10.75 a. -.013 to .093 b. H : Pl - p 2 = 0;
Hi- Pi ~ P2 > 0; critical value: z = 2.33 test statistic:
z = 1.75; do not reject H a
10.77 a. -$119.16 to -$114.84 b. H : p x - p 2 = 0;
H x : p x — p 2 < 0; critical value: z = —1.96; test statistic:
z = -106.25; reject H a
10.79 a. -.086 to .160 fatalities b. H : p x - p 2 = 0;
H x : p x — p 2 > 0; critical value: r = 2.326; test statistic:
/ = .70; do not reject H
10.81 a. -8.42 to -1.82 cards b. H : p x - p 2 = 0;
H x : p x — p 2 + 0; critical values: t = —1.645 and 1.645;
test statistic: t = —3.04; reject H
10.83 a. -.085 to. 159 fatalities; H : p x - p 2 = 0; H x : p x - p 2
> 0; critical value: t = 2.326; test statistic: t = .71; do not
reject H b. —.076 to .150 fatalities; H : p x —
p 2 = 0; H x : p x — p 2 > 0; critical value: t = 2.326; test
statistic: t = .76; do not reject H
10.85 a. -8.35 to -1.89 cards; H : p x - p 2 = 0; H x :
P L i ~ P"2 ^ 0; critical value: t = —1.645 and 1.645; test
statistic: t = -3.11; reject H b. -8.55 to -1.69
cards; H : p x — p 2 = 0; H x : p x — p 2 + 0; critical values:
t = -1.645 and 1.645; test statistic: t = -2.93; reject H Q
10.87 a. -9.54 to -.24 b. H : p d = 0; H x :
p d < 0; critical value: t = —2.896; test statistic: / =
— 2.425; do not reject H
10.89 a. Ind-Rep: -.106 to .046; Ind-Dem: .080 to .220; Rep-
Dem: .103 to .257 b. H : p x - p 2 = 0; H x :
Pi ~ P2 ^ 0; critical values: z = —2.58 and 2.58; test
statistic: z = —.77; do not reject H c. p value =
.4412; for a = .01, do not reject H
Answers to Selected Odd-Numbered Exercises and Self-Review Tests AN9
10.91 a. -.142 to -.018 b. H : p t - p 2 = 0;
Hf. P\ — Pi =fc 0; critical values: z = —1.96 and z = 1.96;
test statistic: z = —2.11; reject H
10.93 .2611
10.95 n = 9
10.97 a. n = 545 b. .8708
10.101 a. .3564 b. .0793 c. .0013
Self- Review Test
I. a
3. a. 1.62 to 2.78 b. H : /x, - p 2 = 0; H,:
p x — p 2 > 0; critical value: z = 1.96; test statistic:
z = 9.86; reject H
4. a. -2.72 to -1.88 hours b. H : p t - p 2 = 0;
Hi: fii — fi2< 0; critical value: / = —2.416; test statistic:
t = - 10.997; reject H„
5. a. -2.70 to -1.90 hours b. H : /a, - p 2 = 0;
H^ p x — p 2 < 0; critical value: f = —2.421; test statistic:
? = -11.474; reject H
6. a. -$53.60 to $186.18 b. H : p d = 0; H,: p d + 0;
critical values: t = —2.447 and 2.447; test statistic:
t = 2.050; do not reject H u
7. a. -.052 to .092 b. H : Pl - p 2 = 0;
H^ Pi ~ P 2 0; critical values: z = —2.58 and 2.58;
test statistic: z = .60; do not reject H
Chapter 11
II. 3 x 1 = 41.337 11.5 x 2 = 41.638
11.7 a. x 2 = 5.009 b. X 2 = 3.565
11.13 critical value: x 2 = 1 1-070; test statistic: x 2 = 5.200; do
not reject H ( ,
11.15 critical value: x 2 = 7.815; test statistic: x 2 = 45.844;
reject H
11.17 critical value: x 2 = 13.277; test statistic: x 2 = 19.328;
reject H
11.19 critical value: x 2 = 9.488; test statistic: x 2 = 6.752; do not
reject H u
11.21 critical value: x 2 = 9.348; test statistic: x 2 = 65.087; reject
H
11.27 a. H : the proportion in each row is the same for all four
populations;
H,: the proportion in each row is not the same for all four
populations
c. critical value: x 2 = 14.449 d. test statistic:
X 2 = 52.451 e. reject H
11.29 critical value: x 2 = 5.024; test statistic: x 2 = 1-980; do not
reject H
11.31 a. critical value: x 2 = 6.635; test statistic: x 2 = 8.647;
reject H b. critical value: x 2 = 6.635; test statistic:
x 2 = 17.317; reject H
11.33 critical value: x 2 = 7.815; test statistic: x 2 = 2.587; do not
reject H
11.35 critical value: x 2 = 6.635; test statistic: x 2 = 8.178; reject H
11.37 critical value: x 2 = 7.815; test statistic: x 2 = 8.221;
reject H
11.39 critical value: x 2 = 7.378; test statistic: x 2 = 2.404; do not
reject H
11.41 a. 18.4376 to 84.9686 b. 21.3393 to 67.7365
c. 23.0674 to 60.6586
11.43 a. H : a 2 = 1.75; H,: a 2 > 1.75
b. reject H„ if x 2 > 34.170
c. test statistic: x 2 = 22.514
d. do not reject H
11.45 a. H : a 2 = 2.2; H{. tr + 2.2 b. reject H if
X 2 < 7.564 or x 2 > 30.191 c. test statistic:
X 2 = 35.545 d. reject H
11.47 a. .8120 to 3.3160; .9011 to 1.8210 b. H : a 2 < 1.0;
Hi. a 2 > 1.0; critical value: x 2 = 41.638; test statistic:
X 2 = 33.81; do not reject H
11.49 a. 2739.3051 to 12,623.9126; 52.338 to 112.356
b. H : a 2 = 4200; H x : v 2 + 4200; critical values:
X 2 = 12.401 and 39.364; test statistic: x 2 = 29.714; do not
11.51
reject H
critical value:
o
X~
= 7.815; test statistic: x 2 =
= 10.464;
11.53
reject H
critical value:
X
= 11.143; test statistic: x 1
= 22.359;
11.55
reject H
critical value:
x 2
= 11.345; test statistic: x 1
= 15.920;
11.57
reject H
critical value:
x 2
= 13,277; test statistic: x 1
= 50.355;
11.59
reject H
critical value:
x~
= 4.605; test statistic: x 2 =
= 13.593;
11.61
reject H
critical value:
not reject H
x~
= 16.812; test statistic: x 2
= 10.181; do
11.63 a. 3.4064 to 24.0000; 1.846 to 4.899 b. 8.3336 to
33.2628; 2.887 to 5.767
11.65 H : o- 2 = 1.1; H{. a 2 > 1.1; critical value: x 2 = 28.845;
test statistic: x 2 = 24.727; do not reject H
11.67 H : o- 2 = 10.4; H,: a 2 + 10.4; critical values: x 2 = 7.564
and 30.191; test statistic: x 1 = 24.192; do not reject H
11.69 a. H : a 2 = 5000; H x : a 2 < 5000; critical value:
X 2 = 8.907; test statistic: x 1 = 12.065; do not reject H„
b. 1666.8509 to 7903.1835; 40.827 to 88.900
11.71 a. .1001 to .4613; .316 to .679 b. H„: a 2 = .13; H,:
tr + .13; critical values: x 2 = 9.886 and 45.559; test
statistic: x 2 = 35.077; do not reject H
11.73 a. s 2 = 1840.6964 b. 804.6509 to 7624.1864;
28.366 to 87.317 c. H : a 2 = 750; H,: a 2 + 750;
critical values: x 2 = 1-690 and 16.013; test statistic:
X 2 = 17.180; reject H
11.75 critical value: x 2 = 5.991; test statistic: x 2 = 12.931;
reject H
11.77 critical value: x 2 = 9.488; test statistic: x 2 = 6.857;
do not reject H
11.79 critical value: x 2 = 16.919; test statistic: x 2 = 215.568;
reject H
11.81 a. test statistic: x 2 = 2.480; p- value > .10 b. no
Self-Review Test
1. b 2. a 3. c 4. a 5. b 6. b
7. c 8. b 9. a
10. critical value: x 2 = 1 1-070; test statistic: x 2 = 20.146; reject
H
11. critical value: x 2 = 11.345; test statistic: x 2 = 31.188; reject
H
12. critical value: x 2 = 9.488; test statistic: x 2 = 82.450; reject H
13. a. .2364 to 1.3326; .486 to 1.154 b. H„: a 2 = .25;
H x : a 2 > .25; critical value: x 2 = 36.191; test statistic:
X 2 = 36.480; reject H
AN 1 Answers to Selected Odd-Numbered Exercises and Self-Review Tests
Chapter 12
12.3
12.5
12.7
12.13
12.15
12.17
12.19
12.21
12.23
12.25
12.27
12.29
12.33
a. 7.26 b. 5.82 c. 5.27
a. 9.00 b. 2.59 c. 1.79
a. 9.96 b. 6.57 12.9 a. 4.85 b. 3.22
a. x t = 15; x 2 = 11; s, = 4.50924975; s 2 = 4.39696865
b. H : yii, = /jl 2 ; H{. (l 1 + /j, 2 ; critical values: / = —2.179
and 2.179; test statistic: t = 1.680; do not reject H
c. critical value: F = 4.75; test statistic: F = 2.82; do not
reject H a d. conclusions are the same
b. critical value: F = 3.29; test statistic: F = 4.07; reject
H
a. H : /x, = (i 2 = yLi 3 = /x 4 ; Hi. all four population means
are not equal b. numerator: df = 3; denominator:
df = 28 c. SSB =.0105; SSW = 1.1449; SST =
1.1554 d. reject H if F > 4.58 e. MSB =
.0035; MSW = .0409 f. critical value: F = 4.58
g. test statistic: F = .0856 i. do not reject H a
critical value: F = 3.55; test statistic: F = 2.09; do not
reject H
critical value: F = 3.72; test statistic: F = 5.44; reject H
a. critical value: F = 2.05; test statistic: F = 2.12;
reject H a
.10
F =
6.93; test statistic: F = 1.24; do not
a. critical value:
reject H
a. critical value: F = 3.89; test statistic: F = 4.89; reject
H b. do not reject H
critical value: F is 5.29; test statistic: F =.57; do not
reject H
a. 5 groups with 10 members each. b. 36 members
each.
Self- Review Test
1. a 2. b 3.
6. a 7. b 8.
10. a. critical value:
b. Type I error
Chapter 13
13.15
4.
F = 3.10; test statistic: F = 4.46; reject H
13.17
13.19
13.21
13.23
13.25
13.27
13.29
13.35
13.37
13.39
a. y-intercept = 100; slope = 5; positive relationship
b. y-intercept = 400; slope = —4; negative relationship
fju ylx = -5.5815 + .2886*
y = -83.7140 + 10.5714*
a. $70.00 b. the same amount
c. exact relationship
a. $27.10 million b. different amounts
c. nonexact relationship
b. y = 322.4483 - 34.4425* e. 8135.10
f. -$29,751.72
b. y = 191.6238 - 25.3714* e. 112.9724
f. -62.0905
a. /x v | v = 630.6627 + 1.2289* b. population
regression line because data set includes all 16 National
League teams; values of A and B d. 734 runs
<r e = 7.0756; p 2 = .04
s e = 4.7117; r 2 = .99
a. SS„ = 64; SS VV = 93636.8889; SS XV = 2283.8
b. s e = 41.6463 c. SST = 93636.8889; SSE =
12140.9133; SSR = 81495.9756 d. r 2 = .87
13.41
a.
s e = 31.2410
13.43
a.
s e = .7832
13.45
a.
<r e = 68.1073
13.47
a.
6.01 to 6.63
13.51
13.53
13.55
13.57
13.63
13.69
13.71
13.73
13.75
13.77
13.79
13.81
13.83
13.85
b. r 2 = .45
b. r = .70
b. p 2 = .02
b. H : B
value: t = 2.145; test statistic: t = 59.792; reject H
c. H„: B = 0; H t : B # 0; critical values: t = -2.977
and 2.977; test statistic: t = 59.792; reject H d. H :
B = 4.50; H x : B + 4.50; critical values: t = -2.624 and
2.624; test statistic: t = 17.219; reject H
13.49 a. 2.35 to 2.65
H : B = 0; H x : B > 0; critical
value: t = 1.960; test statistic: t = 39.124; reject H ;
c. H„: B = 0; H t : B + 0; critical values: t = -2.576
and 2.576; test statistic: t = 39.124; reject H Q \
d. H : B < 1.75; H,: B > 1.75; critical value:
t = 2.326; test statistic: t = 11.737; reject H u
a. -40.3095 to -28.5756 b. H : B = 0; H{. B < 0;
critical value: t = -1.943; test statistic: t = -14.3654;
reject H
a. y = 2.4377* + 25.5536 b. 1.331 to 3.5443
c. H : B = 0; H x : B > 0; critical value: t = 2.365; test
statistic: t = 6.6042; reject H
b. H Q : B = 0; H{. B < 0;
a. -30.5005 to -20.2424
critical value: t = -2.764; test statistic: t = -15.6759;
reject H
a. y = 4.4300 + 1.1403* b. .7041 to 1.5765
c. H : B = 1.0; Hi. B + 1.0; critical values: t = -2.228
and 2.228; test statistic: t = .72; do not reject H a
a 13.67 a. positive b. positive
c. positive d. negative e. zero
p = .21
a. r = -.996 b. H : p = 0; H t : p < 0; critical
value: t = -2.764; test statistic: t = -35.249; reject H
a. positively b. r = .93 c. H : p = 0;
Hi. p > 0; critical value: t = 1.895; test statistic:
t = 6.694; reject H
a. positively b. close to 1 c. r = .97
d. H : p = 0; H { : p + 0; critical values: t = —2.776 and
2.776; test statistic: t = 7.980; reject H
a. r = .93 b. H : p = 0; Hi. p + 0; critical values:
t = -3.499 and 3.499; test statistic: t = 6.694; reject H
p = .16
a. SS a = 750; SS n = 2502.6667; SS n , = 710
b. y = 24.8 + .9467* d. r = .52; i 2 = .27
f. $93.91 g. s e = 13.5297 h. -.15 to 2.05
i. H : B = 0; H{. B > 0; critical value: t = 1.812; test
statistic: t = 1.916; reject H j. H : p = 0; Hf.
p > 0; critical value: t = 2.228; test statistic:
t = 1.925; do not reject H
a.
b.
e.
H h
SS r ,
6394.9; SS m = 1718.9; SS„
y = -22.5355 + .4904* d.
s. = 4.7557 f. .291 to .690
= 3136.1
= .95; r =
g. H :B
.89
= 0;
B > 0; critical value: t = 2.896; test statistic: t = 8.246;
reject H h. H : p = 0; //,: p ¥= 0; critical values:
t = -3.355 and 3.355; test statistic: t = 8.605; reject H
a. SS_„ = 3.3647; SS, V = 788; SS„, = 49.4
b. y = 2.8562 + 14.6819* d. r = .96;
r 2 = .92 e. s, = 3.5416 f. 9.718 to 19.646
g. H : B = 0; H{. B + 0; critical values:
t = -4.032 and 4.032; test statistic: t = 7.6043;
reject H h. H : p = 0; Hi. p > 0; critical value:
/ = 3.365; test statistic: t = 7.6665; reject H
Answers to Selected Odd-Numbered Exercises and Self-Review Tests AN1 1
13.87 a. 13.8708 to 16.6292; 11.7648 to 18.7352
b. 62.3590 to 67.7210; 56.3623 to 73.7177
13.89 $4611.38 to $5374.78; $3808.78 to $6177.38
13.91 $77.63 to $95.03; $54.96 to $117.71
13.93 $1518.85 to $2212.88; $715.60 to $3016.13
13.95 a. positive relationship
b. y = -1.9172 + .9895^: d. r = .97; r 2 = .94
e. s e = 1.0941 f. .54 to 1.44
g. H : B = 0; H x : B > 0; critical value: / = 2.571; test
statistic: t = 8.808; reject H h. H : p = 0; Hy. p > 0;
critical value: t = 2.571; test statistic: f = 8.922; reject
H ; same conclusion
13.97 a. positive b. y = 7.8304 + .5039*
d. r = .89; r 2 = .79 e. 2547 f. s e = 3.3525
g. .11 to .90 h. H : B = 0; H t : B > 0; critical
value: t = 3.365; test statistic: t = 4.278; reject H
i. H : p = 0; H^. p # 0; critical values: t = —3.365 and
3.365; test statistic: t = 4.365; reject H
13.99 a. 55„ = 144.32; SS yy = 3.02329; SS„ = -1.356
b. slightly yes c. y = -.2133 - .0094.x
e. r = -.06 f. -$.29
13.101 b. S xr = 82.5; SS„, =.88956; SS„, = -3.84
c. yes d. y = 22.1615 - .0465.x f. r = -.45
g. 21.65 seconds
13.103 60.7339 to 97.3729; 40.0144 to 118.0924
13.105 1042.7081 to 1153.1345; 953.3648 to 1242.4778
13.107 a. yes b. 246.4670 to 275.5330 lines
c. 200.0567 to 321.9433 lines e. 338 lines
13.111 a. increase b. decrease c. increase
in + 1
d. ±t-s^ —
13.113 a. r = .92; yes
Self-Review Test
1. d 2. a 3. b 4. a 5. b 6. b
7. true 8. true 9. a 10. b
15. a. The attendence depends on temperature.
b. positive d. y = -2.2269 + .2715x
f. r = .65; r 2 = .42 g. 1407 people
h. s e = 3.6172 i. -.30 to .84
j. H : B = 0; H{. B > 0; critical value: t = 3.365;
test statistic: / = 1.904; do not reject H
k. 1055 to 1758 I. 412 to 2401
m. H : p = 0; H{. p > 0; critical value: t = 3.365;
test statistic: / = 1.913; do not reject H
Appendix A
A.7 simple random sample A.9 a. nonrandom sample
b. judgment sample c. selection error
A.ll a. random sample b. simple random sample
c. no
A.13 a. nonrandom sample b. voluntary response error
and selection error
A.15 response error
A.17 a. designed experiment b. no; would need to know if
the women or the doctors who evaluated their health knew
which women took aspirin and which were in the control group
A.19 a. designed experiment b. double-blind study
A.21 designed experiment A.23 yes
A.25 b. observational study c. not a double-blind study
A.27 a. designed experiment b. double-blind study
A.29 a. no b. no c. convenience sample
A.33 a. no b. nonresponse error and response error
c. above
Photo Credits
Chapter 1
Page 1: © WoodyStock/Alamy. Page 16: PhotoDisc,
Inc. /Getty Images.
Chapter 2
Page 27: © Blend Images/SuperStock. Page 29:
PhotoDisc, Inc./Getty Images. Page 43: PhotoDisc,
Inc./Getty Images. Page 56: Mark Harmel/Stone/Getty
Images.
Chapter 3
Page 79: © Martin Thomas Photography/Alamy.
Page 94: Corbis Digital Stock. Page 107: PhotoDisc,
Inc./Getty Images.
Chapter 4
Page 137: Amy Shearer. Page 145: PhotoDisc,
Inc./Getty Images. Page 150: PhotoDisc, Inc./Getty
Images. Page 171: PhotoDisc, Inc./Getty Images.
Chapter 5
Page 191: Martin Diebel /Getty Images, Inc. Page 204:
Corbis Digital Stock. Page 218: PhotoDisc, Inc./Getty
Images.
Chapter 6
Page 250: © Mauritius/SuperStock. Page 274: PhotoDisc,
Inc./Getty Images. Page 276: Courtesy Texas Instru-
ments, Inc.
Chapter 7
Page 300: iStockphoto. Page 308: PhotoDisc, Inc./Getty
Images. Page 317: Corbis Digital Stock. Page 328: Image
State.
Chapter 8
Page 340: © Ted Pink/Alamy. Page 346: © Corbis
Digital Stock. Page 351: PhotoDisc, Inc./Getty Images.
Page 358: PhotoDisc, Inc./Getty Images.
Chapter 9
Page 381: SuperStock. Page 394: Nova Stock/Photo Re-
searchers. Page 408: Cohen/Ostrow/Digital Vision/Getty
Images. Page 409: PhotoDisc, Inc./Getty Images.
Chapter 10
Page 439: © MELBA PHOTO AGENCY/Alamy.
Page 449: PhotoDisc, Inc./Getty Images. Page 452:
Corbis Digital Stock. Page 475: PhotoDisc, Inc./Getty
Images.
Chapter 1 1
Page 498: iStockphoto. Page 504: PhotoDisc, Inc./Getty
Images. Page 516: PhotoDisc, Inc./Getty Images.
Page 527: Corbis Digital Stock.
Chapter 12
Page 541: Steven W. Jones/Taxi/Getty Images. Page 546:
PhotoDisc, Inc./Getty Images. Page 550: PhotoDisc,
Inc./Getty Images.
Chapter 13
Page 564: NBAE/Getty Images, Inc.
Page 589: PhotoDisc, Inc./Getty Images.
Page 599: PhotoDisc, Inc./Getty Images.
Chapter 14
Page 624: PhotoDisc, Inc./Getty Images.
Chapter 15
Page 625: Stephen Marks/The Image Bank/Getty Images.
Appendix A
Page Al: Laura Lane/Taxi/Getty Images.
PCI
A
addition rule, 172-176
defined, 172
for mutually exclusive events, 174-176
in probability of union of events, 172-174
for three mutually exclusive events, 175-176
alternative hypotheses, 382-383
defined, 382, 383
left-tailed test, 387
right-tailed test, 388
test of homogeneity, 517
test of independence, 512
testing, 443
two-tailed test, 386
analysis of variance (ANOVA), 541-563
assumptions, 545
between-samples sum of squares (SSB), 546,
549, 550
defined, 544
degrees of freedom, 549, 550
F distribution, 542-543
mean square between samples (MSB), 545, 547
mean square within samples (MSW), 545, 547
one-way, 544-552
samples, 545
tables, 548, 551
technology instruction, 562-563
test statistic F for, 545, 549, 550-551
total sum of squares (SST), 546, 547, 583
two-way, 545
Type I/Type II error probabilities, 544
within-samples sum of squares (SSW), 546,
550, 551
applied statistics, 2
arithmetic mean. See mean
arrangements. See permutations
average. See mean
B
bar graphs, 3 1
binomial probability distribution, 220, 222, 223
in Minitab, 76
Poisson probability distribution, 235
probability distribution of, 196
bell-shaped distributions
defined, 107
empirical rule, 107-109
normal distribution, 257
Bernoulli trials, 214
between-samples sum of squares (SSB)
calculating, 550-551
defined, 546
formula, substituting values in, 551
bias, 331
bimodal distribution, 86
binomial distribution, 214-224
bar graph, 220, 222, 223
binomial formula and, 216-220
defined, 214, 216
formula, 283
mean of, 223-224
normal approximation to, 283-288
parameters, 216
Poisson distribution as approximation to,
232-233
probability of failure, 214
probability of success, 214, 222-223
shape of, 222-223
standard deviation of, 223-224
table, 220-222
technology instruction, 248
binomial experiments, 214-216
condition verification, 215-216
conditions, 214-215
defined, 214
outcomes, 214
probabilities, 220-222
probability of success, 321
trials, 214
binomial formula
calculating probability with, 216-219
defined, 216
values, substituting, 218
binomial probabilities table, C-2-10
box-and-whisker plots
constructing, 116-117
defined, 115
lower/upper inner fences, 116
12 Index
box-and-whisker plots (continued)
lower/upper outer fences, 117
whiskers, 116
calibration, 331
case studies
Aces High lottery game, 204-205
aggressive driving, 7
anti-bacterial soaps, A- 13
Ask Mr. Statistics, 233
average compensation for accountants, 454—455
baseball player slumps and streaks, 167
births and deaths, 236
candidate decision after interview, 32
career choices for high school students, 31
costs for raising a child, 349
favorite season, 508
favorite seat in the plane, 420
gender pay gap, 85
how crashes affect auto premiums, 399
morning grooming, 40
playing lotto, 212
regression of heights/weights of NBA players, 574-575
rolling stops, 153
sound most frustrating to hear, 365
standard deviation, 108
time distribution to run road race, 255-256
TV commercials, 4
U.S. patent leaders, 3
vacation importance, 478^179
categorical variables. See qualitative variables
causality, 610
census
conducting, 341
defined, 6, 341, A-2
central limit theorem
defined, 313
for sample proportion, 325
central tendency. See measures of central tendency
chance variables. See random variables
Chebyshev's theorem, 106-107, 206
applying, 107
defined, 106
chi-square distribution, 499-501
curves, 499
defined, 498, 499
degrees of freedom, 499
mean, 499
symbol, 498
table, C-23
table, reading, 499-501
chi-square tests, 498-540
goodness-of-fit, 502-509
population variance, 523-528
p-value, 509
technology instruction, 539-540
test of homogeneity, 517-519
test of independence, 512-517
classes
boundaries, 35
defined, 35
less-than method, 41^-3
lower limit, 35, 37-38
midpoint, 36
number of, 37
single-valued, 43-44
size, 36
upper limit, 35
width, 36, 37
classical probability rule, 144-145
cluster sampling, A-9
clusters, A-9
coefficient of determination, 582-585
calculating, 585
concept, 582
defined, 584, 585
formula, 584-585
coefficient of variation, 98
coefficient of x, 566
combinations, 209-211
defined, 210
finding, 210-211
formula, 210, 211
notation, 210
number of, 210
technology instruction, 248
combined mean, 90
complementary events
calculating probabilities of, 157-158
defined, 156-157
Venn diagram, 157
compound events
calculating probability of, 144-145
defined, 140
illustrating, 141-142
conditional probability
calculating, 152-153, 164-165
defined, 151
required, 53, 152
confidence coefficients, 343
confidence intervals
confidence level and width of, 348
defined, 343
difference between two population means, 442-443, 448^149,
458^159
difference between two population proportions, 474-475
mean of population paired differences, 465-467
population mean, 345
population mean with t distribution, 357-359
population proportion, 363
population variance, 524-525
regression line slope, 588
sample size of, 348-350
technology instruction, 378-379
width of, 348-350, 359
confidence levels
defined, 343
denotation, 343
interpretation, 347
width of confidence interval and, 348
z values for, 346
consistent estimator, 308, 325
constant term, 567
Index 13
contingency tables, 511
continuity correction factor
defined, 285
using, 286, 287, 288
continuous probability distribution, 251-254
characteristics, 252
probability calculation, 253
single value, 253, 254
continuous random variables. See also normal distribution
defined, 193, 251
examples, 193
probability distribution curves, 252
single value probability, 253, 254
values, 250
control groups, A-3, 11
controlled experiments, A- 10
convenience samples, A-5
correction factors
continuity, 285, 286, 287, 288
finite population, 307, 325
correction for continuity, 285
defined, 285
making, 286, 287, 288
correlation coefficient
defined, 564
linear, 592-595
outliers and, 611-612
square of, 594
counting rule, 149-150
critical values (points)
defined, 383
of F, 549, 550
of t, 409, 410, 590
of z, 418, 419
critical-value approach. See also hypothesis tests
defined, 389, 394-395
left-tailed test (population mean), 397-399
left-tailed test (population proportion), 419^121
performance steps, 395
population mean (population standard deviation known),
390-394, 394-399
population mean (population standard deviation not known),
408-411
population proportion, 417^-21
right-tailed test (population mean), 409-410
test statistic, 395
two-tailed test (population mean), 395-397, 408^109
two-tailed test (population proportion), 417—419
cross-section data, 13
cross-tabulation. See contingency tables
cumulative frequency distributions. See also frequency
distributions
defined, 51
ogive, 52-53
percentages, 52
relative frequencies, 52
D
data
cross-section, 13
defined, 9
entering and saving, 22-26
grouped, 35
organizing, 75-78
primary, A-2
qualitative, 28-33
quantitative, 35^-4
raw, 28
secondary, A-2
sources, 14-15, A-l-3
time-series, 13-14
ungrouped, 35, 80
data sets
bimodal, 86
defined, 3, 9
explanation of, B-l-4
multimodal, 86
population, 96
sample, 96
unimodal, 86
decision making. A- 9- 10
degrees of freedom
chi-square distribution, 499
defined, 355
F distribution, 542, 543
goodness-of-fit test, 503
number of, 355
one-way ANOVA, 549, 550
paired samples, 464
population variance, 526, 527
prediction interval, 609
for simple linear regression model, 581
t distribution, 355, 458
test of independence, 512
of two samples taken together, 448
dependent events, 155, 156
dependent samples, 440
dependent variables, 565
descriptive statistics, 3
designed experiments, A-ll-13. See also experiments
control group, A- 11
defined, A- 11
double-blind. A- 12
example, A- 11— 12
observational study versus, A- 11
placebo effect, A- 12
treatment group, A- 1 1
deterministic models, 567
difference between two population means, independent samples,
440^161
confidence interval, 442-443, 448-449, 458-459
hypothesis testing, 443-445, 450^53, 459^161
p- value approach, 445, 451, 453, 461
sample size and, 453
standard deviation known, 440-445
standard deviation unknown and unequal, 457-461
standard deviation unknown but equal, 447^-53, 457-461
difference between two population means, paired samples,
464^171
hypothesis testing, 467-471
interval estimation, 465-467
left-tailed test of hypothesis, 467-469
p- value approach, 469, 471
two-tailed test of hypothesis, 469-471
14 Index
difference between two population proportions, 473-481
hypothesis testing, 476-481
interval estimation, 474^175
p- value approach, 477^178, 480-481
right-tailed test of hypothesis, 476-478
two-tailed test of hypothesis, 478^181
difference between two sample means
mean, 440-442
sampling distribution, 440-442
standard deviation, 440^142
test statistic, 444, 445, 450, 451, 460, 461
difference between two sample proportions
mean, 474
sampling distribution, 474
standard deviation, 474
test statistic, 476, 477, 480
discrete random variables, 192-193
defined, 192
dichotomous, 214
examples, 192-193
finding probabilities of events for, 197
mean of, 201-202
occurrences, 230
probability distribution, 194-198, 206
standard deviation of, 202-206
dispersion. See measures of dispersion
distributions
bell-shaped, 107-109, 257
bimodal, 86
binomial, 214-224
chi-square, 499-501
continuous, 251-254
cumulative frequency, 51, 52-53
exponential, 290
F, 542-543, C-24-27
frequency, 28-30, 30-33, 35-36, 51-53
hypergeometric, 226-229
memoryless, 290
multimodal, 86
normal, 254-258
percentage, 30, 38-39, 52
Poisson, 230-235
population, 301, 311, 313-314, 545
sampling, 300, 465, 474, 523, 587
sampling, of p, 323-330
sampling, of x, 301-339
shape, 296
standard normal, 259-265
unimodal, 86
dotplots, 58-60
defined, 58
in Minitab, 77
stacked, 59
uses, 58, 59
double-blind experiments, A-3, 12
E
educated guesses, 2
elementary events. See simple events
elements
defined, 3
of sample space, 138
of samples, 8
empirical rule, 107-109
defined, 107
illustrated, 108
equally likely outcomes, 144
equation of linear relationship, 566
error sum of squares (SSE), 570
errors
hypothesis test, 384-385
nonresponse, A-7
nonsampling, 303-305, A-6-8
of prediction, 572
random, 570, 581-582
response, A-7
sampling, 303, 304-305, A-5-6
selection, A-6-7
standard deviation of, 581-582
total, 583
Type I, 384-385
Type II, 385
voluntary, A-7-8
estimated regression model, 568
estimates
of A and B, 568
defined, 341, 342
point, 342
of standard deviation, 458
estimation
defined, 341
interval, 342-343
introduction to, 341-342
least squares regression line, 571-572
of population mean, 344-35 1
procedure, 342
of regression line slope, 588
two populations, 439^197
estimators
consistent, 308, 325
defined, 307, 324, 341, 342
of standard deviation of difference between two sample
means, 448
unbiased, 307, 324
events
complementary, 156-158
compound, 140-141
dependent, 155, 156
equally likely, 144
impossible, 143
independent, 155-156, 165-167
intersection of, 161-162
mutually exclusive, 154-155, 167-168
mutually nonexclusive, 154
probability of, 143
simple, 140, 141
sure, 143
union of, 171-172
exact relationship, deterministic model, 567
Excel
analysis of variance (ANOVA), 562-563
chi-square tests, 540
Index 15
column creation, 26
combinations, binomial distribution, and Poisson
distribution, 248
confidence intervals, 379
entering/saving data, 25-26
hypothesis testing, 436-437
Minitab, 134-136
normal and inverse normal probabilities, 297
organizing data, 78
random number generation, 189-190
sampling distribution of means, 339
simple linear regression, 621-622
sum of columns, 26
two populations, 494-495
expected frequencies. See also frequencies
defined, 502, 503
test of homogeneity, 519
test of independence, 512-514
expected values, 201-202
experiments
binomial, 214-216
controlled, A- 10
as data source, A-3
defined, 138, A-3
designed, A- 11-1 3
double-blind, A-3, 12
final outcomes for, 143
multinomial, 502
random (chance), 192
exponential distribution, 290
external data sources, A-2
external sources, 14
extrapolation, 609-610
extreme outliers, 117
extreme values. See outliers
F distribution, 542-543
characteristics, 542
curves, 542
defined, 542
degrees of freedom, 542, 543
table, C-24-27
table, reading, 543
factorials, 208-209
defined, 209
evaluating, 209
symbol, 208
finite population correction factor, 307, 325
first quartile, 1 10
formulation, 424
frequencies
defined, 35
expected, 502, 503
joint, 511
observed, 502, 530
frequency distribution curves
defined, 41
mean, median and mode, 87
skewed, 45, 87
symmetric, 45
frequency distribution tables
classes, 35-38
constructing, 29-30, 36-38
cumulative, 51
defined, 29
grouped data, 35
less-than method, 41^43
for qualitative data, 29-30
for quantitative data, 35-38
single-valued classes, 43^-4
frequency distributions
cumulative, 51-53
defined, 29
for qualitative data, 28-30
for quantitative data, 35-36
of sample mean, 302, 303
of sample proportion, 324
frequency histograms, 39, 40
frequency polygons, 40-41
geometric mean, 91
goodness-of-fit test, 502-509. See also chi-square tests
conducting, 504-509
defined, 502
degrees of freedom, 503
equal proportions, 504-506
expected frequencies, 502, 503
observed frequencies, 502, 503
results fit given distribution, 506-509
as right-tailed test, 503
sample size, 504
test statistic, 503, 505-506, 507
government publications, 14
graphing
axes, truncating, 61-62
Poisson probability distribution, 235
qualitative data, 31-33
quantitative data, 39-44
grouped data
basic formulas for variance and standard deviation,
128
defined, 35
mean for, 99-101
standard deviation for, 101-103
variance for, 101-103
histograms
defined, 39
frequency, 39, 40
median, 85
normal approximation to the binomial, 284
percentage, 39
relative frequency, 40, 41
shapes of, 44-46
skewed, 45
symmetric, 44-45
uniform or rectangle, 45
homogeneity, test of. See test of homogeneity
16 Index
hypergeometric probability distribution, 226-229
defined, 227
formula, 227, 228, 229
probability calculation with, 227-229
hypotheses
alternative, 382-383
null, 382-383
rejection/nonrejection regions, 383-384
hypothesis tests
approaches to, 389
critical-value approach, 389
difference between two population means, 443-445, 450^153,
459^461
difference between two population proportions, 476-481
error types, 384-385
goodness-of-fit, 502-509
homogeneity, 517-519
independence, 512-517
left-tailed, 386, 387
linear correlation, 595-596
mean of population paired differences, 467-471
not significantly different, 400
population mean (population standard deviation known),
390^100
population mean (population standard deviation not known),
404-411
population proportion, 414-421
population variance, 525-528
power of, 385
p-value approach, 389
regression line slope, 588-590
rejection and nonrejection regions, 383-384
right-tailed, 386, 387-388
significance level, 384
significantly different, 400
tails of, 386-389
technology instruction, 434^-37
two populations, 439^-97
two-tailed, 386-387
I
impossible events, 143
independence, test of. See test of independence
independent (explanatory) variables, 565
independent events
calculating joint probability of, 165-166
defined, 155
illustrating, 155-156
multiplication rule for, 165-167
independent samples
defined, 440
difference between two population means, 440-461
difference between two population proportions,
473-481
example, 440
inferential statistics, 3-4, 340
internal data sources, A-l
interquartile range (IQR), 111, 112
intersection of events, 161-162
defined, 161
illustrated, 162, 163
interval estimation, 342-344
confidence level, 343
defined, 342
lower limit, 343
margin of error, 343
upper limit, 343
J
joint frequencies, 511
joint probability, 162-168
calculating, 162-165
in conditional probability calculation, 164-165
defined, 162
of independent events, 165-166
multiplication rule for finding, 162-167
of mutually exclusive events, 167-168
tree diagram for, 163, 164, 166
judgment samples, A-5
L
Law of Large Numbers, 146
least squares estimates of A and B, 571
least squares method, 569
least squares regression line, 569-572
defined, 569, 570
error sum of squares (SSE), 570
estimating, 571-572
observed (actual) value of y, 569
predicted value of y, 569
left-tailed test, 386, 387
critical-value approach (population mean), 397-399
critical-value approach (population proportion), 419-421
mean of population paired differences, 467-469
p-value for, 394, 407^108
less-than method (classes), 41-43
line graphs. See bar graphs
linear correlation, 592-595
between two variables, 593
calculating, 594
defined, 592
formula, 593-594
hypothesis testing, 595-596
perfect positive, 593
p-value approach, 596
strong negative, 593
strong positive, 593
test statistic, 595, 596
value of, 593
weak negative, 593
weak positive, 593
lineal' regression model, 565-567
defined, 565
equation of linear relationship, 566
illustrated, 566
lists
entering data in, 22
establishing, 22-23
names, changing, 22-23
numeric operations on, 23
lower inner fences, 116
lower outer fences, 117
Index 17
margin of error
defined, 343, 345
of estimate of population mean, 350
of estimate of population proportion, 363, 364
interval estimation of difference between two population
means, 443
size, predetermining, 350
marginal probability, 150-151
matched samples. See paired samples
mean
of binomial distribution, 223-224
calculating, 80, 99-101
chi-square distribution, 499
combined, 90
defined, 80
difference between two sample means, 440^142
difference between two sample proportions, 474
of discrete random variables, 201-202
frequency distribution curves, 87
geometric, 91
for grouped data, 99-101
of normal distribution, 257, 258, 269, 285
outlier sensitivity, 82, 83
paired difference, 465
of Poisson distribution, 235
population, 99-100
regression line slope, 587
relationships with median and mode, 86-87
sample, 100-101
of sample mean, 306-307
of sample proportion, 324, 328, 330
of sampling distribution of p, 324-325
of t distribution, 355
for ungrouped data, 80-83
weighted, 91
of x, 306-308, 311-312, 318
mean of population paired differences
hypothesis testing, 467-471
inferences about, 465
interval estimation, 465^167
left-tailed test of hypothesis, 467-469
p- value approach, 469, 471
mean of sample paired differences
mean, 465
sampling distribution, 465
standard deviation, 465
test statistic, 467, 468-469
mean square between samples (MSB)
in ANOVA table, 548
defined, 545
formula, 547
ratio, 548, 551
mean variance within samples (MSW)
in ANOVA table, 548
defined, 545
formula, 547
ratio, 548, 551
measurements. See observations
measures of central tendency
defined, 80
mean, 80-83
median, 83-85
mode, 85-86
relationships among mean, median, and mode, 86-87
for ungrouped data, 80-87
measures of dispersion
coefficient of variation, 98
defined, 92
population parameters, 96
range, 92-93
sample statistics, 96
standard deviation, 93-96
variance, 93-96
measures of position
defined, 110
interquartile range, 110-112
percentile rank, 113-114
percentiles, 113
quartiles, 110-112
median
calculating, 83
defined, 83
frequency distribution curves, 87
as histogram center, 85
outliers and, 85
relationships with mean and mode, 86-87
members. See elements
memoryless distribution, 290
mild outliers, 1 17
Minitab
analysis of variance (ANOVA), 562
bar charts, 76
chi-square tests, 539
column sum calculation, 24-25
columns, creating, 24
combinations, binomial distribution, and Poisson distribution, 248
confidence intervals, 378
dotplot, 77
entering and saving data, 23-24
Excel, 134-136
hypothesis testing, 435-436
normal and inverse normal probabilities, 296-297
numerical descriptive measures, 132-134
pie charts, 76-77
random number generation, 189
sampling distribution of means, 338-339
simple linear regression, 621
stem-and-leaf display, 77
two populations, 491-494
mode
calculating, 85, 86
defined, 85
frequency distribution curves, 87
relationships with mean and median, 86-87
shortcoming, 86
multimodal distribution, 86
multinomial experiments, 502
multiple regression, 565
multiplication rule, 162-167
of independent events, 165-167
for three events, 166-167
18 Index
mutually exclusive events, 154-155
addition rule for, 174-176
defined, 154
illustrating, 154-155
joint probability of, 167-168
probability of union, 174-176
Venn diagram of, 174
mutually nonexclusive events, 154
N
negative linear relationship, 573
nonlinear regression model, 565-566
nonrandom samples, A-4— 5
nonrejection region. See rejection/nonrejection regions
nonresponse errors, A-7
nonsampling errors. See also errors
defined, 303, A-6
examples, 304-305
minimizing, 304
nonresponse, A-7
occurrence, 304, A-6
response, A-7
selection, A-6-7
voluntary response, A-7-8
normal approximation to the binomial, 283-288
correction for continuity, 286, 287
defined, 283-284
formula, 383
histogram, 284
using, 286-288
normal distribution, 254-258
applications of, 273-276
bell-shaped curve, 257
center of, 258
characteristics, 257
defined, 107, 257
different means, same standard deviation, 258
equation of, 258
family of curves, 258
finding x value for, 280-282
mean of, 257, 258, 269, 285
one-tailed test with, 393-394
parameters, 258
probability of points left of mean, 275
probability of points right of mean, 275-276
same mean, different standard deviation, 258
standard, 259-265
standard deviation of, 257, 258, 285
standardizing, 267-272
symmetric about the mean, 257, 384
t distribution and, 355
tails, 257-258
technology instruction, 296-297
total area under, 257
two-tailed test with, 392-393
use of, 254
X and z values for known area, 278-282
x value conversion to z value, 267
normal random variables, 257
null hypotheses, 382-383
defined, 382, 383
left-tailed test, 387
rejecting, 397, 400
rejection/nonrejection regions, 383-384
right-tailed test, 388
test of homogeneity, 517
test of independence, 512
two-tailed test, 386
numerical descriptive measures, 79-136
central tendency for grouped data, 98-105
central tendency for ungrouped data, 80-87
dispersion for grouped data, 98-105
dispersion for ungrouped data, 92-96
mean, 80-83, 99-101
median, 83-85
Minitab, 132-134
mode, 85-87
position, 110-114
standard deviation, 93-96, 101-103, 105-110
technology instruction, 132-136
variance, 93-96, 101-103
observational studies
defined, A- 10
designed experiments versus, A- 11
example, A-l 1
variable relationship, A- 12
observational units. See elements
observations
defined, 3, 9
number of, 38
observed frequencies
defined, 502, 503
test of independence, 512, 513
observed values
of sample mean, 396
of test statistic for sample mean, 405
of z value, 392, 397, 414
occurrences. See also Poisson probability distribution
average number of, 23 1
defined, 230
odds, probability and, 179-180, 239-240
ogives, 52-53
one-way ANOVA, 544-552. See also analysis of variance (ANOVA)
assumptions, 545
defined, 544
degrees of freedom, 549, 550
samples, 545
tables, 548, 551
test, 548-552
test preparation, 548-549
test statistic F for, 545, 549, 550-551
outcomes
binomial experiment, 214
defined, 138
equally likely, 144
final, 143
total, counting rule to find, 149-150
outliers
correlation and, 611-612
defined, 82
Index 19
detecting, 58
extreme, 117
mean and, 82, 83
median and, 85
mild, 117
paired difference
defined, 464
hypothesis testing, 467-471
interval estimation, 465^167
mean, 465
population, 465^4-7 1
sample, 465
standard deviation, 465
values, 464
paired samples
defined, 440
degrees of freedom, 464
difference between two population means, 464-471
sample size, 464
pairwise comparisons, 561
parameters
binomial, 216
normal distribution, 258
Poisson, 231
population, 96, 301-302, 568
percentage distribution
cumulative, 52
defined, 30
for qualitative data, 30
for quantitative data, 38-39
percentage histograms, 39
percentage polygons, 41
percentile rank, 113-114
percentiles, 113
permutations, 212-213
defined, 213
formula, 213
notation, 213
pie charts, 32-33
angle size calculation, 32
defined, 32
in Minitab, 76-77
for percentage distribution, 32-33
placebo effect, A- 12
point estimates
defined, 342
finding for population proportion, 363-364
Poisson probabilities table, C-13-18
Poisson probability distribution, 230-235
application examples, 230-231
as approximation to binomial distribution,
232-233
average number of occurrences, 231
bar graph, 235
conditions to apply, 230
constructing, 235
defined, 230
formula, 231, 232
graphing, 235
intervals, 232
mean of, 235
occurrences, 230
parameter of, 23 1
standard deviation of, 235
table, 233-235
technology instruction, 248
polygons, 40^-1
pooled sample proportion, 476
pooled standard deviation, 448
population distribution
defined, 301
illustrated, 311, 313, 314
variance of, 545
population means
calculating, 99-100
confidence interval, 345
difference between two (independent samples, known), 440-453
difference between two (independent samples, unknown and
unequal), 457-461
difference between two (independent samples, unknown but
equal), 447^153
difference between two (paired samples), 464^-71
estimation (population standard deviation known), 344-35 1
estimation (population standard deviation unknown), 354-359
estimator of, 341
hypothesis tests (population standard deviation known),
390-400
hypothesis tests (population standard deviation not known),
404-411
t distribution and, 357-359
true, 341
two, confidence interval, 442-443, 448-449
two, hypothesis testing, 443-445, 450-453
population parameters
defined, 96
estimating, 341
point estimates, 342
in probabilistic models, 568
values, 301-302
values, assignment of, 341
population proportions, 321-322
calculation of, 322
confidence interval, 363
confidence interval, finding, 363-364
critical-value approach, 417-421
defined, 322
denotation, 362
difference between two, 473^-81
estimation (large samples), 362-367
estimation (most conservative), 366
estimation (preliminary sample results), 366-367
estimator of, 341
estimator of standard deviation, 363
hypothesis tests, 414^121
left-tailed test (critical-value approach), 419^-21
margin of error, 363, 364
observed value of z, 414
point estimate, 363-364
p- value approach, 414-417
right-tailed test (p-value approach), 416-417
110 Index
population proportions {continued)
sample size determination, 364-367
true, 341
two-tailed test, (p-value approach), 414-416
two-tailed test (critical-value approach), 417-419
population regression line
defined, 568
error distribution around, 576
formula, 571
population standard deviation, 93
hypothesis tests and, 390 — 4 1 1
known, 344-351, 390-400
obtaining, 94, 101
population mean estimation and, 344-359
unknown, 354-359, 404-411
population variance. See also chi-square tests
confidence interval, 524-525
degrees of freedom, 526, 527
hypothesis tests, 525-528
inferences about, 523-528
right-tailed test, 526-527
test statistic, 526, 527, 528
two-tailed test, 527-528
populations
defined, 3
sample proportions and, 321-322
samples versus, 5-8
sampling (normally distributed), 310-312
sampling (not normally distributed), 313-314
sampling distribution, 301-303
subpopulations, A-9
target, 5
position. See measures of position
positive linear relationship, 573
power of the test, 385
prediction interval, 608-609
defined, 608
making, 609
primary data, A-2
primary units, A-9
probabilistic models, 567-568
defined, 567
population parameters, 568
random error term, 568, 574, 576
probability, 137-190
addition rule, 172-176
binomial distribution success, 222-223
binomial experiment, 220-222
calculating, 143-147
classical, 144-145
complementary events, 156-158
compound event calculation, 144-145
conceptual approaches, 144-147
conditional, 151-153
continuous probability distribution, 253
counting rule, 149-150
defined, 4, 137, 143
dependent events, 155, 156
with hypergeometric distribution formula, 227-229
independent events, 155-156
intersection of events, 161-162
joint, 162-168
marginal, 150-151
multiplication rule and, 162-167
mutually exclusive events, 154-155
odds and, 179-180, 239-240
with Poisson formula, 231-233
properties, 143
relative frequency concept of, 145-147
simple event calculation, 144
statistics and, 179
subjective, 147
union of events, 171-172
probability density functions, 252
probability distribution
binomial, 214-224
characteristics, 195
constructing, 198
continuous, 251-254
defined, 194
discrete random variables, 194-198, 206
graphical presentation of, 196
hypergeometric, 226-229
Poisson, 230-235
presentation, 195-196
tree diagram, 198
two conditions, 195
verifying conditions of, 196
probability distribution curve
in case study, 256
of continuous random variables, 252
total area under, 252
probability-value approach. See p-value approach
processing errors, 611
proportions. See population proportions; sample proportions
p-value approach. See also hypothesis tests
chi-square test, 509
defined, 389, 391
difference between two population means, 445, 451, 453, 461
difference between two population proportions, 477-478,
480-481
for left-tailed test (population mean), 394, 407-408
linear correlation, 596
mean of population paired differences, 469, 471
performance steps, 392
population mean (population standard deviation known), 391-394
population mean (population standard deviation not known), 405
population proportion, 414—417
p-value calculation, 406, 407-408, 415, 416, 416^117
range for, 405
in regression analysis, 602-604
regression line slope, 590
right-tailed test (population mean), 391
right-tailed test (population proportion), 416—417
two-tailed test (population mean), 391, 393, 405-406
two-tailed test (population proportion), 414^116
qualitative data
frequency distributions, 28-30
graphing, 30-33
histograms, 39^-0
Index 111
organizing, 28-30
percentage distribution, 30
polygons, 40^4-1
relative frequency, 30
qualitative variables, 12
quantitative data
cumulative frequency distributions, 51-53
cumulative percentages, 52
cumulative relative frequencies, 52
dotplots, 58-60
frequency distribution tables, 36-38
frequency distributions, 35-36, 41-44
graphing, 39^14
ogives, 52-53
organizing, 35-39
percentage distributions, 38-39
relative frequency, 38-39
stem-and-leaf displays, 54—55
quantitative variables, 1 1
quartiles
defined, 110
finding, 111-112
first, 110
position of, 110-111
second, 110
third, 110
quota samples, A-5
random (chance) experiments, 192
random errors, 570, 574. See also errors
for sample regression model, 570
spread of, 581
standard deviation of, 581-582, 601
sum of, 570
random number generation, 189-190
random samples
defined, 6, A-4
as representative samples, A-4
simple, 6
random sampling techniques, A-8-9
random variables. See also variables
binomial, 224
continuous, 195
defined, 192
discrete, 192-193, 194-198, 201-206
sample variance as, 523
randomization, A- 10
range
calculating, 92-93
defined, 92
disadvantages, 93
raw data, 28
rectangular histograms, 45
regression line slope, 566
confidence interval of, 588
hypothesis testing, 588-590
inferences about, 587-590
mean, 587
p-value approach, 590
sampling distribution, 587
standard deviation, 587-588
test statistic, 589
true values of, 568
y-intercept and, 567
regression lines. See also simple linear regression
coefficient of x, 566
estimating with samples, 608
finding, 600
least squares, 569-573
population, 568, 576
true values of y-intercept and slope, 568
y-intercept, 566
regression models. See also simple linear regression
assumptions, 574-576
defined, 565
degrees of freedom, 581
deterministic, 567
equation of, 568
estimated, 568
estimated (predicated) value of y, 568
estimated values of A and B, 568
for estimating mean value of y, 606-608
exact relationship, 567
linear, 565, 566
nonlinear, 565, 566
population parameters, 568
for predicting particular value of y, 608-609
probabilistic, 567-568
random errors, 570, 581-582
statistical relationship, 567-568
using, 606-609
regression of y on x, 568
regression sum of squares (SSR), 584
rejection/nonrejection regions, 383-384
defined, 383-384
difference between two population means, 450, 452, 460^-61
difference between two population proportions, 477, 480
goodness-of-fit test, 504-505, 507
linear correlation, 595-596
mean of population paired differences, 468, 470
one-way ANOVA test, 549, 550
population mean (standard deviation known), 396, 398
population mean (standard deviation not known), 409, 410
population proportion, 418, 419
population variance, 516, 527-528
in regression analysis, 602, 603
regression line slope, 589
test of homogeneity, 518-519
test of independence, 514-515, 516
relative frequencies. See also frequencies
cumulative, 52
defined, 30
for qualitative data, 30
for quantitative data, 38-39
of sample mean, 302, 303
of sample proportion, 324
relative frequency concept of probability, 145-147
defined, 145
Law of Large Numbers and, 146
using, 145-147
relative frequency densities, 252
112 Index
relative frequency histograms, 40, 41
relative frequency polygons, 41
representative samples, 6
residual, 569
response errors, A-7
right-tailed test, 386, 387-388
critical-value approach, 409-410
difference between two population means, 452^-53
difference between two population proportions, 476-478
goodness-of-fit test as, 503
population variance, 526-527
p- value for (population mean), 391
p- value for (population proportion), 416-417
test of homogeneity as, 518
test of independence as, 514
s
sample means
central limit theorem for, 313
defined, 302
frequency distribution of, 303
mean of, 306-307
observed value of, 396
probability, as interval, 317-319
relative frequency distribution of, 303
sampling distribution of, 302, 303
sampling error and, 305
shape of, 310-311
standard deviation of, 306-307
two, difference between, 440^-42
z value for, 392
sample points, 138
sample proportions, 321-322
applications of, 328-330
calculation of, 322, 323
central limit theorem of, 325
as consistent estimator, 325
defined, 322
as estimator, 324
estimator of standard deviation of, 363
frequency distribution of, 324
mean of, 324, 328, 330
pooled, 476
probability, a certain value, 329-330
probability, an interval, 328-329
relative frequency distribution of, 324
sampling distribution of, 323
standard deviation of, 324-325, 328, 330
as unbiased estimator, 324
z value for, 329
sample size
difference between two population means, independent
samples and, 453
for estimation of mean, 350-35 1
for estimation of population proportion, 364-367
goodness-of-fit test, 504
paired samples, 464
sample proportion and, 324
t distribution and, 359, 410-411
test of independence, 5 14
width of confidence interval and, 348-350
sample space, 138
sample standard deviation
obtaining, 94, 101
pooled, 448
sample statistics, 96
sample surveys, 6, A-2
sample variance, 523
samples
ANOVA, 545
convenience, A-5
defined, 3, 5
dependent, 440
elements, 8
independent, 440-461, 473^181
judgment, A-5
mean, calculating, 100-101
nonrandom, A-4, 5
paired (matched), 440, 464^171
populations versus, 5-8
quota, A-5
random, 6, A-4
representative, 6
variance between, 545
variance within, 545
sampling
random, A-8-9
reasons for, A-4
with replacement, 6-8
without replacement, 8
sampling distribution, 587
defined, 300
difference between two sample proportions, 474
of mean of sample paired difference, 465
of regression line slope, 587
of sample variance, 523
sampling distribution of p, 323-330
applications of, 328-330
defined, 323
example, 323-324
mean of, 324-325
p value, 323
sample size and, 324
shape of, 325-326
standard deviation of, 324-325
sampling distribution of x, 301-339
applications of, 316-319
central limit theorem and, 313
difference between two sample means, 440-442
mean of, 306-308, 311-312, 318
normally distributed population, 310-312
not normally distributed population, 313-314
observations, 308
population and, 301-303, 311, 313
probability calculation, 317-319
shape of, 310-314
spread of, 308
standard deviation of, 306-308, 311-312, 318
technology instruction, 337-339
x value, 317
sampling errors, A-5-6
defined, 303, A-5
Index 113
examples, 304-305
occurrence, 303, A-6
scatter diagrams, 568-569, 600-601
defined, 569
illustrated, 569
second quartile, 110
secondary data, A-2
selection errors, A-6-7
short-cut formulas
for standard deviation, 94, 101
for variance, 94, 101
£ (sigma) notation, 15-16
significance level
defined, 343, 384
regression, 602
simple events
calculating probability of, 144
defined, 140
illustrating, 141-142
sum of probabilities of, 143
simple linear regression, 564-623. See also regression lines;
regression models
analysis, 567-577, 599-604
causality and, 610
cautions, 609-610
coefficient of determination, 582-585
coefficient of x, 566
confidence interval for B, 588, 601
dependent variables, 565
equation of linear relationship, 566
extrapolation and, 609-610
hypothesis testing, 588-590, 602
independent variables, 565
interpretation of a and b, 573
least squares line, 569-572
linear, 565-567
linear correlation, 592-596
linear regression, 565-567
multiple, 565
negative linear relationship, 573
nonlinear relationship between x and y, 577
observed (actual) value of y, 569
positive linear relationship, 573
predicted value of y, 569
p-value approach, 602-604
random error term, 568, 574, 576
random errors, 570, 581-582
regression of y on x, 568
regression sum of squares (SSR), 584
scatter diagram, 568-569, 600-601
significance level, 602
simple, 565
simple regression, 565
standard deviation of errors, 581-582, 601
technology instruction, 620-622
test statistic, 602, 603
understanding uses/limitations, 620
y-intercept, 566
simple probability, 150-151
simple random samples, 6
simple random sampling, 6, A-8
simple regression, 565
single-valued classes, 43-44
skewed histograms, 45
slope. See regression line slope
sources, data, 14-15, A- 1—3
experiments, A-3
external, A-2
internal, A-l
surveys, A-2-3
specification, 424
SSE. See error sum of squares
SSR. See regression sum of squares
SST. See total sum of squares
standard deviation
basic formulas, 93
basic formulas (grouped data), 128
basic formulas (ungrouped data), 127
of binomial distribution, 223-224
calculating, 94-96, 101-103
case study, 108
Chebyshev's theorem, 106-107, 206
defined, 93
difference between two sample means, 440^-42, 448
difference between two sample proportions, 474
empirical rule, 107-109
estimate of, 458
estimator of sample proportion, 363
for grouped data, 101-103
of mean of sample paired difference, 465
of normal distribution, 257, 258, 285
obtaining, 93
of paired differences, 465
of Poisson distribution, 235
pooled, 448
population, 93, 94, 101, 344-351
of random errors, 581-582, 601
regression line slope, 587-588
sample, 93, 94, 101
of sample mean, 306-307
of sample proportion, 324-325, 328, 330
of sampling distribution of p, 324-325
short-cut formulas for, 94, 101
of t distribution, 355
for ungrouped data, 93-96, 94
use of, 105-109
values, 93, 95
of I, 306-308, 311-312, 318
standard deviation of discrete random variable, 202-206
calculating, 203-206
defined, 202
formula, 202
interpretation of, 206
standard error of x, 306
standard normal distribution, 259-265
defined, 259
table, 264, C- 19-20
z values (z scores), 259, 260, 261, 262, 263
standard normal distribution curve
area under, 259
defined, 259
one standard deviation of the mean, 264
114 Index
standard normal distribution curve {continued)
three standard deviations of the mean, 264
two standard deviations of the mean, 264
standard units (standard scores). See z values
standardizing normal distributions, 267-272
defined, 267
x value conversion to z value, 267
Statistical Abstract of the United States, 14
statistical properties, 74—75
statistical relationship, probabilistic model, 567-568
statistics
applied, 2
defined, 2
descriptive, 3
inferential, 3^1, 340
language of, 18
probability and, 179
theoretical, 2
types of, 2-4
stem-and-leaf displays, 54-56
construction procedure, 54-55
defined, 54
grouped, 56
in Minitab, 77
ranked, 55-56
strata, A-9
stratified random sampling, A-8-9
strong negative linear correlation, 593
strong positive linear correlation, 593
Student's t distribution. See t distribution
subjective probability, 147
subpopulations, A-9
summation notation, 15-17
defined, 15
one variable, 16
two variables, 16-17
sure events, 143
surveys
census, A-2
conducting, A- 2-3
defined, 6, A-2
sample, 6, A-2, 4-5
symmetric histograms, 44-45
systematic random sampling, A-8
t distribution
confidence interval for population mean with, 357-359
defined, 355
degrees of freedom, 355, 458
mean of, 355
normal distribution and, 355
sample size and, 359, 41CMH1
standard deviation of, 355
symmetric shape, 356, 357
table, 356, 358, 409, 410, C-21-22
tables
ANOVA, 548
binomial probabilities, C-2-10
binomial probability distribution, 220-222
chi-square distribution, C-23
contingency, 511
F distribution, C-24-27
frequency distribution, 29-38, 41-44
Poisson distribution, 233-235
Poisson probabilities, C-13-18
standard normal distribution, C- 19-20
t distribution, 356, 358, 409, 410, C-21-22
values of e~\ C-ll-12
tails, test, 385-389
left-tailed test, 386, 387
right-tailed test, 386, 387-388
two-tailed test, 386-387
target population, 5
technology instruction
analysis of variance (ANOVA), 562-563
chi-square tests, 539-540
combinations, binomial distribution, and Poisson distribution, 248
confidence intervals, 378-379
entering and saving data, 22-26
hypothesis tests, 434-437
normal and inverse normal probabilities, 296-297
numerical descriptive measures, 132-136
organizing data, 75-78
random number generation, 189-190
sampling distribution of means, 337-339
simple linear regression, 620-622
two populations, 491^-95
test of homogeneity, 517-519
alternative hypothesis, 517
defined, 517
expected frequencies, 519
making, 518-519
null hypothesis, 517
as right-tailed test, 518
test statistic, 519
test of independence, 512-517
alternative hypothesis, 512
defined, 512
degrees of freedom, 512
expected frequencies, 512-514
null hypothesis, 512
observed frequencies, 512, 513
as right-tailed test, 514
sample size, 514
test statistic for, 512, 515, 516-517
2X2 table, 515-517
2X3 table, 514-515
test statistic
calculating value of, 396-397, 398, 409, 410, 418, 420
computed value of, 397
critical values of, 409, 410
defined, 395, 405
difference between two sample means, 444, 445, 450, 451,
460, 461
difference between two sample proportions, 476, 477, 480
goodness-of-fit test, 503, 505-506, 507
linear correlation, 595, 596
mean of sample paired differences, 467, 470
observed value of, 405
for one-way ANOVA test, 545, 549, 550-551
population variance, 526, 527, 528
Index 115
in regression analysis, 602, 603
regression line slope, 589
test of homogeneity, 519
test of independence, 512, 515, 516-517
value of, 396
theoretical statistics, 2
third quartile, 110
TI-84
analysis of variance (ANOVA), 562
changing list names/establishing lists, 22-23
chi-square tests, 539
combinations, binomial distribution, and Poisson distribution, 248
confidence intervals, 378
data organization, 75-76
entering data in lists, 22
hypothesis testing, 434
normal and inverse normal probabilities, 296
numeric operations on lists, 23
numerical descriptive measures, 132
random number generation, 189
sampling distribution of means, 337-338
simple linear regression, 620
two populations, 491
time-series data, 13-14
total errors, 583
total sum of squares (SST), 546, 547
defined, 583
ratio, 584
traditional (classical) approach. See critical-value approach
treatment
defined, A- 10
groups, A-3, 1 1
tree diagrams
defined, 138
drawing, 138-140
illustrated, 139, 140
for joint probabilities, 163, 164, 166
probability distribution, 198
probability of union of three mutually exclusive events, 176
trials, 214
true population mean, 341
true population proportion, 341
two populations
means, independent samples, 440^161
means, paired samples, 464^-71
proportions, 473-481
technology instruction, 491-495
two-tailed test, 386-387
critical value approach (population proportion), 417-419
critical-value approach (population mean), 395-397, 408^-09
difference between two population means, 444-445,
450-451, 460^161
difference between two population proportions, 478-481
mean of population paired differences, 469-471
population variance, 527-528
p-value (population mean), 391, 393, 405^-06
p-value (population proportion), 414^4-16
two-way ANOVA, 545
Type I errors, 384-385, 544
Type II errors, 385, 544
typical values, 80
u
unbiased estimator, 307, 324
ungrouped data
basic formulas for variance and standard deviation, 127
defined, 35, 80
mean for, 80-83
measures of central tendency for, 80-87
measures of dispersion, 92-96
median for, 83-85
mode for, 85-86
range for, 92-93
standard deviation for, 93-96
variance for, 93-96
uniform histograms, 45
unimodal distribution, 86
union of events, 171-176
calculating, 172-174
defined, 171
illustrating, 171-172
mutually exclusive events, 174-176
upper inner fences, 116
upper outer fences, 117
uses and misuses
bias, 331
coin flips, 424
exponential distribution, 290
formulation (specification), 424
language of statistics, 18
national versus local unemployment rate, 370
odds and probabilities, 179-180, 239-240
on-time airline performance, 555
outliers and correlation, 611-612
population differences, 483
processing errors, 611
putting on game face, 238-239
statistics versus probability, 179
truncating the axes, 61-62
unemployment rates, 118-119
wildlife habits, 530
V
variables
continuous, 1 1-12
defined, 8
dependent, 565
discrete, 11
independent (explanatory), 565
linear correlation between, 593
negative relationship between, 573
positive relationship between, 573
qualitative, 12
quantitative, 11
random, 192-193, 194-198, 201-206
relationships, 564
types of, 10-12
variance
analysis of (ANOVA), 541-563
basic formulas (grouped data), 128
basic formulas (ungrouped data), 127
between samples, 545
calculating, 94-96, 101-103
116 Index
variance (continued)
defined, 93
for grouped data, 101-103
measurement units of, 95
population, 523-528, 545
sample, 523
short-cut formulas for, 94, 101
for ungrouped data, 93-96
values, 95
within samples, 545
Venn diagrams
complementary events, 157
defined, 138
drawing, 138-140
illustrated, 139, 140, 141
mutually exclusive events, 174
voluntary response errors, A-7-8
w
weak negative linear correlation, 593
weak positive linear correlation, 593
weighted mean, 91
whiskers, 116
width of confidence interval, 348-350, 359
within-samples sum of squares (SSW)
calculating, 550-551
defined, 546
formulas, substituting values in, 551
X
x
coefficient of, 566
equation of linear relationship, 566
positive/negative linear relationships, 573
regression of y on, 568
in simple linear regression analysis, 567
x values
conversion to z values, 267
determining, 278-282
finding for normal distribution, 280-282
Y
y
equation of linear relationship, 566
mean value, estimating, 606-607
nonlinear relationship between x and, 577
observed (actual) value of, 569
positive/negative linear relationships, 573
predicted value of, 569
regression on x, 568
value, predicting, 608-609
y-intercept
defined, 566
true values of, 568
z
z values
in confidence interval formula, 345
for confidence levels, 346
defined, 259
determining, 278-282
negative, 259, 261, 263
observed value of, 392, 397, 414
positive, 259, 260, 262
for sample mean, 392
standard normal distribution table and, 345
test statistic, 414, 418
for value of p, 329
for value of x, 317
x value conversion to, 267
Table IV Standard Normal Distribution Table
The entries in this table give the
cumulative area under the standard
normal curve to the left of z with the
values of z equal to or negative.
z
.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
-3.4
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0003
.0002
-3.3
.0005
.0005
.0005
.0004
.0004
.0004
.0004
.0004
.0004
.0003
-3.2
.0007
.0007
.0006
.0006
.0006
.0006
.0006
.0005
.0005
.0005
-3.1
.0010
.0009
.0009
.0009
.0008
.0008
.0008
.0008
.0007
.0007
-3.0
.0013
.0013
.0013
.0012
.0012
.0011
.0011
.0011
.0010
.0010
-2.9
.0019
.0018
.0018
.0017
.0016
.0016
.0015
.0015
.0014
.0014
-2.8
.0026
.0025
.0024
.0023
.0023
.0022
.0021
.0021
.0020
.0019
-2.7
.0035
.0034
.0033
.0032
.0031
.0030
.0029
.0028
.0027
.0026
-2.6
.0047
.0045
.0044
.0043
.0041
.0040
.0039
.0038
.0037
.0036
-2.5
.0062
.0060
.0059
.0057
.0055
.0054
.0052
.0051
.0049
.0048
-2.4
.0082
.0080
.0078
.0075
.0073
.0071
.0069
.0068
.0066
.0064
-2.3
.0107
.0104
.0102
.0099
.0096
.0094
.0091
.0089
.0087
.0084
-2.2
.0139
.0136
.0132
.0129
.0125
.0122
.0119
.0116
.0113
.0110
-2.1
.0179
.0174
.0170
.0166
.0162
.0158
.0154
.0150
.0146
.0143
-2.0
.0228
.0222
.0217
.0212
.0207
.0202
.0197
.0192
.0188
.0183
-1.9
.0287
.0281
.0274
.0268
.0262
.0256
.0250
.0244
.0239
.0233
-1.8
.0359
.0351
.0344
.0336
.0329
.0322
.0314
.0307
.0301
.0294
-1.7
.0446
.0436
.0427
.0418
.0409
.0401
.0392
.0384
.0375
.0367
-1.6
.0548
.0537
.0526
.0516
.0505
.0495
.0485
.0475
.0465
.0455
-1.5
.0668
.0655
.0643
.0630
.0618
.0606
.0594
.0582
.0571
.0559
-1.4
.0808
.0793
.0778
.0764
.0749
.0735
.0721
.0708
.0694
.0681
-1.3
.0968
.0951
.0934
.0918
.0901
.0885
.0869
.0853
.0838
.0823
-1.2
.1151
.1131
.1112
.1093
.1075
.1056
.1038
.1020
.1003
.0985
-1.1
.1357
.1335
.1314
.1292
.1271
.1251
.1230
.1210
.1190
.1170
-1.0
.1587
.1562
.1539
.1515
.1492
.1469
.1446
.1423
.1401
.1379
-0.9
.1841
.1814
.1788
.1762
.1736
.1711
.1685
.1660
.1635
.1611
-0.8
.2119
.2090
.2061
.2033
.2005
.1977
.1949
.1922
.1894
.1867
-0.7
.2420
.2389
.2358
.2327
.2296
.2266
.2236
.2206
.2177
.2148
-0.6
.2743
.2709
.2676
.2643
.2611
.2578
.2546
.2514
.2483
.2451
-0.5
.3085
.3050
.3015
.2981
.2946
.2912
.2877
.2843
.2810
.2776
-0.4
.3446
.3409
.3372
.3336
.3300
.3264
.3228
.3192
.3156
.3121
-0.3
.3821
.3783
.3745
.3707
.3669
.3632
.3594
.3557
.3520
.3483
-0.2
.4207
.4168
.4129
.4090
.4052
.4013
.3974
.3936
.3897
.3859
-0.1
.4602
.4562
.4522
.4483
.4443
.4404
.4364
.4325
.4286
.4247
0.0
.5000
.4960
.4920
.4880
.4840
.4801
.4761
.4721
.4681
.4641
Table IV Standard Normal Distribution Table (continued)
The entries in this table give the
cumulative area under the standard
normal curve to the left of z with the
values of z equal to or positive.
z
.00
.01
.02
.03
.04
.05
.06
.07
.08
.09
0.0
.5000
.5040
.5080
.5120
.5160
.5199
.5239
.5279
.5319
.5359
1
U. 1
5308
5438
5478
1 7
— Ul I
5557
.JJ7U
5636
5675
.5714
5753
— 1 1 J J
9
u.z
5703
5839
5871
.JO / 1
5010
.J71U
5048
5087
.J70 /
6096
.uuzu
6064
.6103
.6141
3
U.J
.6179
.6217
6955
6703
6331
6368
6406
6443
6480
6517
.UJ 1 1
0.4
.6554
.6591
.6628
.6664
.6700
.6736
.6772
.6808
.6844
.6879
0.5
.6915
.6950
.6985
.7019
.7054
.7088
.7123
.7157
.7190
.7224
n 6
u.u
7957
. / zj /
7901
7394
. / Jit
7357
7380
. / joy
7499
. / HZZ
7454
7486
75 1 7
7540
. / Jt7
7
u. /
7580
.761 1
.7642
7673
. / u / J
.7704
.7734
7764
7704
. / / y^t
7873
. / OZ J
7859
u.o
7881
. / oo 1
70i n
. / y iu
7030
. / 7J7
7067
. / 7U /
7005
. / 77J
8093
.ouzo
805 1
8078
.ou /
8 1 06
.0 1UU
81 33
0.9
.8159
.8186
.8212
.8238
.8264
.8289
.8315
.8340
.8365
.8389
1.0
.8413
.8438
.8461
.8485
.8508
.8531
.8554
.8577
.8599
.8621
1.1
8643
86 65
8686
.ouou
8708
.o / uo
8790
.O / i7
8740
.o / ^ry
8770
.0 / / u
8700
.0 / yyj
8810
8830
.OO JU
1.2
8X4.0
88(S0
8888
.oooo
8007
.07U /
8095
8044
8069
.07UZ
8080
8007
.oyy 1
001 5
i . j
0037
.7U
004.0
0066
.7UW
0089
0000
.7W77
.711J
01 31
.71 Jl
01 47
01 69
.7 1UZ
01 77
.y 1 / 1
1.4
.9192
.9207
.9222
.9236
.9251
.9265
.9279
.9292
.9306
.9319
1.5
.9332
.9345
.9357
.9370
.9382
.9394
.9406
.9418
.9429
.9441
1 .o
.y^-Oj
QA1A
.yH 1 h
.7JUj
.yj 1 J
.yjZj
.7JJJ
.7j4j
1 7
.yjj^
.7J / J
.7JOZ
.yjy 1
Q^QQ
.7J77
.7OU0
.70jj
1 S
1 .0
.yon- 1
Q6SA
.7DJD
.7004
QA71
.yO 1 1
.70 / o
.7O0O
.7O7J
.7D77
.7 /UO
1.9
.9713
.9719
.9726
.9732
.9738
.9744
.9750
.9756
.9761
.9767
2.0
.9772
.9778
.9783
.9788
.9793
.9798
.9803
.9808
.9812
.9817
z.l
noo 1
.yozo
.yoj4
nolo
OO/I o
.yo4z
flQ/l A
.yo4o
no cf\
.yoDV
.yo34
.yoj /
Z.Z
QQA 1
.yovl
QQ.6/1
QG6Q.
.yo 1 1
.yo 1 j
QG7Q.
.Vo /o
.Vool
.yoo4
.yoo /
.yoy\J
2.3
.9893
.9896
.9898
.9901
.9904
.9906
.9909
.991 1
.9913
.9916
2.4
.9918
.9920
.9922
.9925
.9927
.9929
.9931
.9932
.9934
.9936
2.5
.9938
.9940
.9941
.9943
.9945
.9946
.9948
.9949
.9951
.9952
2.6
.9953
.9955
.9956
.9957
.9959
.9960
.9961
.9962
.9963
.9964
2.7
.9965
.9966
.9967
.9968
.9969
.9970
.9971
.9972
.9973
.9974
2.8
.9974
.9975
.9976
.9977
.9977
.9978
.9979
.9979
.9980
.9981
2.9
.9981
.9982
.9982
.9983
.9984
.9984
.9985
.9985
.9986
.9986
3.0
.9987
.9987
.9987
.9988
.9988
.9989
.9989
.9989
.9990
.9990
3.1
.9990
.9991
.9991
.9991
.9992
.9992
.9992
.9992
.9993
.9993
3.2
.9993
.9993
.9994
.9994
.9994
.9994
.9994
.9995
.9995
.9995
3.3
.9995
.9995
.9995
.9996
.9996
.9996
.9996
.9996
.9996
.9997
3.4
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9997
.9998
Table V The f Distribution Table
The entries in this table give the critical values
of f for the specified number of degrees
of freedom and areas in the right tail.
Area in the Right Tail under the t Distribution Curve
if
.10
.05
.025
.01
.005
.001
1
3.078
6.314
12.706
31.821
63.657
318 309
2
1.886
2.920
4.303
6.965
9.925
22.327
3
1.638
2.353
3.182
4.541
5.841
10.215
4
1.533
2.132
2.776
3.747
4.604
7.173
5
1.476
2.015
2.571
3.365
4.032
5.893
6
1.440
1.943
2.447
3.143
3.707
5.208
7
1.415
1.895
2.365
2.998
3.499
4.785
8
1.397
1.860
2.306
2.896
3.355
4.501
9
1.383
1.833
2.262
2.821
3.250
4.297
10
1.372
1.812
2.228
2.764
3.169
4.144
11
1.363
1.796
2.201
2.718
3.106
4.025
12
1.356
1.782
2.179
2.681
3.055
3.930
13
1.350
1.771
2.160
2.650
3.012
3.852
14
1.345
1.761
2.145
2.624
2.977
3.787
15
1.341
1.753
2.131
2.602
2.947
3.733
16
1.337
1.746
2.120
2.583
2.921
3.686
17
1.333
1.740
2.110
2.567
2.898
3.646
18
1.330
1.734
2.101
2.552
2.878
3.610
19
1.328
1.729
2.093
2.539
2.861
3.579
20
1.325
1.725
2.086
2.528
2.845
3.552
21
1.323
1.721
2.080
2.518
2.831
3.527
22
1.321
1.717
2.074
2.508
2.819
3.505
Zj
1 T 1 Q
1 "71/1
z.jUU
Z.oU /
24
1.318
1.711
2.064
2.492
2.797
3.467
25
1.316
1.708
2.060
2.485
2.787
3.450
26
1.315
1.706
2.056
2.479
2.779
3.435
27
1.314
1.703
2.052
2.473
2.771
3.421
28
1.313
1.701
2.048
2.467
2.763
3.408
29
1.311
1.699
2.045
2.462
2.756
3.396
30
1.310
1.697
2.042
2.457
2.750
3.385
31
1.309
1.696
2.040
2.453
2.744
3.375
32
1.309
1.694
2.037
2.449
2.738
3.365
33
1.308
1.692
2.035
2.445
2.733
3.356
34
1.307
1.691
2.032
2.441
2.728
3.348
35
1.306
1.690
2.030
2.438
2.724
3.340
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
00
The t Distribution Table (continued)
Area in the Right Tail under the t Distribution Curve
.10
.05
.025
.01
.005
.001
.306
1.688
2.028
2.434
2.719
3.333
.305
1.687
2.026
2.431
2.715
3.326
.304
1.686
2.024
2.429
2.712
3.319
.304
1.085
i mo
2.023
i /i it
2.420
i 7no
2. /05
1 1 1 Q
3.313
.303
1 tO A
1.084
1 AT 1
2.021
1 /I n
2.423
1 Ht\A
2. /04
3.30/
.303
1.683
2.020
2.421
2.701
3.301
.302
1.682
2.018
2.418
2.698
3.296
.302
1.681
2.017
2.416
2.695
3.291
.301
1 ton
1.080
2.U15
2.414
i tni
2.oy2
i lot
3.250
.301
1.0/9
i m a
2.014
1/111
2.412
i tnn
2.09U
1 1 O 1
3.251
.300
1.679
2.013
2.410
2.687
3.277
.300
1.678
2.012
2.408
2.685
3.273
.299
1.677
2.011
2.407
2.682
3.269
inn
1.0/ /
i n 1 n
2.U1U
i a nc
2.405
i ton
2.05U
q itc
3.2o5
inn
1.0/0
i nnn
2.009
i a m
2.403
2.0/5
q it 1
3.2ol
.298
1.675
2.008
2.402
2.676
3.258
.298
1.675
2.007
2.400
2.674
3.255
.298
1.674
2.006
2.399
2.672
3.251
.297
1.674
2.005
2.397
2.670
3.248
.29 /
1.0/3
i nn/i
2.004
i Qn£
2.390
1 tto
2.005
1 1 A C
3.245
.297
1.673
2.003
2.395
2.667
3.242
.297
1.672
2.002
2.394
2.665
3.239
.296
1.672
2.002
2.392
2.663
3.237
int
1 £.H 1
1.0/ 1
i nn 1
2.U01
i in 1
2.391
i tti
2.002
i n a
3.234
on/;
1.0/ 1
i nnn
2.U00
i Qnn
2.390
i ttn
2.000
1 1 Q 1
5. HI
.296
1.670
2.000
2.389
2.659
3.229
.295
1.670
1.999
2.388
2.657
3.227
.295
1.669
1.998
2.387
2.656
3.225
in?
1 ttn
LOW
i nno
1.998
2.350
2.055
5. Hi
.295
1.669
1.997
2.385
1 £C A
2.654
3.220
.295
1.668
1.997
2.384
2.652
3.218
.294
1.668
1.996
2.383
2.651
3.216
.294
1.668
1.995
2.382
2.650
3.214
.294
1.667
1.995
2.382
2.649
3.213
.294
1.667
1.994
2.381
2.648
3.211
.294
1.667
1.994
2.380
2.647
3.209
.293
1.666
1.993
2.379
2.646
3.207
.293
1.666
1.993
2.379
2.645
3.206
.293
1.666
1.993
2.378
2.644
3.204
.293
1.665
1.992
2.377
2.643
3.202
.282
1.645
1.960
2.326
2.576
3.090